Behavioral dynamics of steering, obstacle avoidance, and route selection

Adjustments of speed and path when avoiding collisions with another pedestrian

When walking in open space, collision avoidance with other pedestrians is a process that successfully takes place many times. To pass another pedestrian (an interferer) walking direction, walking speed or both can be adjusted. Currently, the literature is not yet conclusive of how humans adjust these two parameters in the presence of an interferer. This impedes the development of models predicting general obstacle avoidance strategies in humans’ walking behavior. The aim of this study was to investigate the adjustments of path and speed when a pedestrian is crossing a non-reactive human interferer at different angles and speeds, and to compare the results to general model predictions. To do so, we designed an experiment where a pedestrian walked a 12 m distance to reach a goal position. The task was designed in such a way that collision with an interferer would always occur if the pedestrian would not apply a correction of movement path or speed. Results revealed a strong dependence of path and speed adjustments on crossing angle and walking speed, suggesting local planning of the collision avoidance strategy. Crossing at acute angles (i.e. 45° and 90°) seems to require more complex collision avoidance strategies involving both path and speed adjustments than crossing at obtuse angles, where only path adjustments were observed. Overall, the results were incompatible with predictions from existing models of locomotor collision avoidance. The observed initiations of both adjustments suggest a collision avoidance strategy that is temporally controlled. The present study provides a comprehensive picture of human collision avoidance strategies in walking, which can be used to evaluate and adjust existing pedestrian dynamics models, or serve as an empirical basis to develop new models.

Follow the leader: visual control of speed in pedestrian following

When people walk together in groups or crowds they must coordinate their walking speed and direction with their neighbors. This paper investigates how a pedestrian visually controls speed when following a leader on a straight path (one-dimensional following). To model the behavioral dynamics of following, participants in Experiment 1 walked behind a confederate who randomly increased or decreased his walking speed. The data were used to test six models of speed control that used the leader's speed, distance, or combinations of both to regulate the follower's acceleration. To test the optical information used to control speed, participants in Experiment 2 walked behind a virtual moving pole, whose visual angle and binocular disparity were independently manipulated. The results indicate the followers match the speed of the leader, and do so using a visual control law that primarily nulls the leader's optical expansion (change in visual angle), with little influence of change in disparity. This finding has direct applications to understanding the coordination among neighbors in human crowds.

Displaying optic flow to simulate locomotion: Comparing heading and steering

Optic flow can be used by humans to determine their direction of heading as well as controlling steering. Dot-flow displays have been widely used to investigate these abilities but it is unclear whether photorealistic textures would provide better information for controlling high-speed steering. Here, we examine the accuracy of heading judgements from dot-flow displays of different densities and luminance and then compare to a scene containing a textured ground. We then examine steering behaviour using these same displays to determine whether accurate heading conditions necessarily equate to successful steering. Our findings suggest that the bright dense dot-flow displays led to equivalent performance as the ground texture when judging heading, and this was also true when steering. The intermediate dot-flow conditions (with fewer and faded dots) revealed that some conditions that led to accurate heading judgements were insufficient for accurate steering. It seems, therefore, that heading perception should not be considered synonymous with successful steering control, and displays that support one ability will not necessarily support the other.

Making inferences about intention: perceptual control theory as a "theory of mind" for psychologists

Theory of Mind (ToM) assumes that humans and possibly other primates understand behavior in terms of inferences about intentions. While there is evidence that primates make such inferences, little attention has been paid to the question of their validity. In order to answer this question it is necessary to know the true intentions underlying behavior. The present paper shows that Perceptual Control Theory can provide a scientific basis for making such determinations using methods derived from control engineering. These methods--called the "Test for the Controlled Variable" (TCV)--are based on the assumption that intentional behavior is equivalent to the process of control. The TCV provides an objective approach to inferring the intentions underlying behavior in terms of the perceptual variables under control and the goal states of those variables. Thus, Perceptual Control Theory represents an empirical ToM for psychologists--one that can be used to understand behavior in terms of inferences about intention that are based on the results of active experimentation rather than passive observation.

Blind(fold)ed by science: a constant target-heading angle is used in visual and nonvisual pursuit

Previous work investigating the strategies that observers use to intercept moving targets has shown that observers maintain a constant target-heading angle (CTHA) to achieve interception. Most of this work has concluded or indirectly assumed that vision is necessary to do this. We investigated whether blindfolded pursuers chasing a ball carrier holding a beeping football would utilize the same strategy that sighted observers use to chase a ball carrier. Results confirm that both blindfolded and sighted pursuers use a CTHA strategy in order to intercept targets, whether jogging or walking and irrespective of football experience and path and speed deviations of the ball carrier during the course of the pursuit. This work shows that the mechanisms involved in intercepting moving targets may be designed to use different sensory mechanisms in order to drive behavior that leads to the same end result. This has potential implications for the supramodal representation of motion perception in the human brain.

Modular inverse reinforcement learning for visuomotor behavior

In a large variety of situations one would like to have an expressive and accurate model of observed animal or human behavior. While general purpose mathematical models may capture successfully properties of observed behavior, it is desirable to root models in biological facts. Because of ample empirical evidence for reward-based learning in visuomotor tasks, we use a computational model based on the assumption that the observed agent is balancing the costs and benefits of its behavior to meet its goals. This leads to using the framework of reinforcement learning, which additionally provides well-established algorithms for learning of visuomotor task solutions. To quantify the agent's goals as rewards implicit in the observed behavior, we propose to use inverse reinforcement learning, which quantifies the agent's goals as rewards implicit in the observed behavior. Based on the assumption of a modular cognitive architecture, we introduce a modular inverse reinforcement learning algorithm that estimates the relative reward contributions of the component tasks in navigation, consisting of following a path while avoiding obstacles and approaching targets. It is shown how to recover the component reward weights for individual tasks and that variability in observed trajectories can be explained succinctly through behavioral goals. It is demonstrated through simulations that good estimates can be obtained already with modest amounts of observation data, which in turn allows the prediction of behavior in novel configurations.

Guiding locomotion in complex, dynamic environments

Locomotion in complex, dynamic environments is an integral part of many daily activities, including walking in crowded spaces, driving on busy roadways, and playing sports. Many of the tasks that humans perform in such environments involve interactions with moving objects-that is, they require people to coordinate their own movement with the movements of other objects. A widely adopted framework for research on the detection, avoidance, and interception of moving objects is the bearing angle model, according to which observers move so as to keep the bearing angle of the object constant for interception and varying for obstacle avoidance. The bearing angle model offers a simple, parsimonious account of visual control but has several significant limitations and does not easily scale up to more complex tasks. In this paper, I introduce an alternative account of how humans choose actions and guide locomotion in the presence of moving objects. I show how the new approach addresses the limitations of the bearing angle model and accounts for a variety of behaviors involving moving objects, including (1) choosing whether to pass in front of or behind a moving obstacle, (2) perceiving whether a gap between a pair of moving obstacles is passable, (3) avoiding a collision while passing through single or multiple lanes of traffic, (4) coordinating speed and direction of locomotion during interception, (5) simultaneously intercepting a moving target while avoiding a stationary or moving obstacle, and (6) knowing whether to abandon the chase of a moving target. I also summarize data from recent studies that support the new approach.

Kinematic evaluation of virtual walking trajectories

Virtual walking, a fundamental task in Virtual Reality (VR), is greatly influenced by the locomotion interface being used, by the specificities of input and output devices, and by the way the virtual environment is represented. No matter how virtual walking is controlled, the generation of realistic virtual trajectories is absolutely required for some applications, especially those dedicated to the study of walking behaviors in VR, navigation through virtual places for architecture, rehabilitation and training. Previous studies focused on evaluating the realism of locomotion trajectories have mostly considered the result of the locomotion task (efficiency, accuracy) and its subjective perception (presence, cybersickness). Few focused on the locomotion trajectory itself, but in situation of geometrically constrained task. In this paper, we study the realism of unconstrained trajectories produced during virtual walking by addressing the following question: did the user reach his destination by virtually walking along a trajectory he would have followed in similar real conditions? To this end, we propose a comprehensive evaluation framework consisting on a set of trajectographical criteria and a locomotion model to generate reference trajectories. We consider a simple locomotion task where users walk between two oriented points in space. The travel path is analyzed both geometrically and temporally in comparison to simulated reference trajectories. In addition, we demonstrate the framework over a user study which considered an initial set of common and frequent virtual walking conditions, namely different input devices, output display devices, control laws, and visualization modalities. The study provides insight into the relative contributions of each condition to the overall realism of the resulting virtual trajectories.

Modeling heading and path perception from optic flow in the case of independently moving objects

Humans are usually accurate when estimating heading or path from optic flow, even in the presence of independently moving objects (IMOs) in an otherwise rigid scene. To invoke significant biases in perceived heading, IMOs have to be large and obscure the focus of expansion (FOE) in the image plane, which is the point of approach. For the estimation of path during curvilinear self-motion no significant biases were found in the presence of IMOs. What makes humans robust in their estimation of heading or path using optic flow? We derive analytical models of optic flow for linear and curvilinear self-motion using geometric scene models. Heading biases of a linear least squares method, which builds upon these analytical models, are large, larger than those reported for humans. This motivated us to study segmentation cues that are available from optic flow. We derive models of accretion/deletion, expansion/contraction, acceleration/deceleration, local spatial curvature, and local temporal curvature, to be used as cues to segment an IMO from the background. Integrating these segmentation cues into our method of estimating heading or path now explains human psychophysical data and extends, as well as unifies, previous investigations. Our analysis suggests that various cues available from optic flow help to segment IMOs and, thus, make humans' heading and path perception robust in the presence of such IMOs.

Obstacle avoidance and smooth trajectory control: neural areas highlighted during improved locomotor performance

Visual control of locomotion typically involves both detection of current egomotion as well as anticipation of impending changes in trajectory. To determine if there are distinct neural systems involved in these aspects of steering control we used a slalom paradigm, which required participants to steer around objects in a computer simulated environment using a joystick. In some trials the whole slalom layout was visible (steering “preview” trials) so planning of the trajectory around future waypoints was possible, whereas in other trials the slalom course was only revealed one object at a time (steering “near” trials) so that future planning was restricted. In order to control for any differences in the motor requirements and visual properties between “preview” and “near” trials, we also interleaved control trials which replayed a participants' previous steering trials, with the task being to mimic the observed steering. Behavioral and fMRI results confirmed previous findings of superior parietal lobe (SPL) recruitment during steering trials, with a more extensive parietal and sensorimotor network during steering “preview” compared to steering “near” trials. Correlational analysis of fMRI data with respect to individual behavioral performance revealed that there was increased activation in the SPL in participants who exhibited smoother steering performance. These findings indicate that there is a role for the SPL in encoding path defining targets or obstacles during forward locomotion, which also provides a potential neural underpinning to explain improved steering performance on an individual basis.

Embodied Cognition is Not What you Think it is

The most exciting hypothesis in cognitive science right now is the theory that cognition is embodied. Like all good ideas in cognitive science, however, embodiment immediately came to mean six different things. The most common definitions involve the straight-forward claim that “states of the body modify states of the mind.” However, the implications of embodiment are actually much more radical than this. If cognition can span the brain, body, and the environment, then the “states of mind” of disembodied cognitive science won’t exist to be modified. Cognition will instead be an extended system assembled from a broad array of resources. Taking embodiment seriously therefore requires both new methods and theory. Here we outline four key steps that research programs should follow in order to fully engage with the implications of embodiment. The first step is to conduct a task analysis, which characterizes from a first person perspective the specific task that a perceiving-acting cognitive agent is faced with. The second step is to identify the task-relevant resources the agent has access to in order to solve the task. These resources can span brain, body, and environment. The third step is to identify how the agent can assemble these resources into a system capable of solving the problem at hand. The last step is to test the agent’s performance to confirm that agent is actually using the solution identified in step 3. We explore these steps in more detail with reference to two useful examples (the outfielder problem and the A-not-B error), and introduce how to apply this analysis to the thorny question of language use. Embodied cognition is more than we think it is, and we have the tools we need to realize its full potential.

Dynamical Movement Primitives: Learning Attractor Models for Motor Behaviors

Nonlinear dynamical systems have been used in many disciplines to model complex behaviors, including biological motor control, robotics, perception, economics, traffic prediction, and neuroscience. While often the unexpected emergent behavior of nonlinear systems is the focus of investigations, it is of equal importance to create goal-directed behavior (e.g., stable locomotion from a system of coupled oscillators under perceptual guidance). Modeling goal-directed behavior with nonlinear systems is, however, rather difficult due to the parameter sensitivity of these systems, their complex phase transitions in response to subtle parameter changes, and the difficulty of analyzing and predicting their long-term behavior; intuition and time-consuming parameter tuning play a major role. This letter presents and reviews dynamical movement primitives, a line of research for modeling attractor behaviors of autonomous nonlinear dynamical systems with the help of statistical learning techniques. The essence of our approach is to start with a simple dynamical system, such as a set of linear differential equations, and transform those into a weakly nonlinear system with prescribed attractor dynamics by means of a learnable autonomous forcing term. Both point attractors and limit cycle attractors of almost arbitrary complexity can be generated. We explain the design principle of our approach and evaluate its properties in several example applications in motor control and robotics.

Intercepting a moving traffic gap while avoiding collision with lead and trail vehicles: gap-related and boundary-related influences on drivers' speed regulations during approach to an intersection

Using a fixed-base driving simulator, 15 participants actively drove their vehicle across a rural road toward an intersection. Their task was to safely cross the intersection, passing through a gap in the train of incoming traffic. Spatiotemporal task constraints were manipulated by varying the initial conditions (offsets) with respect to the time of arrival of the traffic gap at the intersection. Orthogonally manipulating the motion characteristics of the lead and trail vehicles forming the traffic gap allowed evaluating the influences of the global (gap-related) and local (lead/trail-vehicle-related) aspects of the inter-vehicular interval. The results revealed that the different initial offsets gave rise to functional, continuous and gradual adjustments in approach speed, initiated early on during approach to the intersection. Drivers systematically accelerated during the final stages of approach, on average crossing the gap slightly ahead of the center of the traffic gap. A special-purpose ANOVA demonstrated an influence of (global) gap characteristics such as gap size and speed. Further analyses demonstrated that the motion characteristics of the lead vehicle exerted a stronger influence on approach behavior than the motion characteristics of the trail vehicle. The results are interpreted as signing the online regulation of approach speed, concurrently based on intercepting the (center of the) traffic gap and avoiding collision with the lead and trail vehicles.

Intersection crossing considered as intercepting a moving traffic gap: effects of task and environmental constraints

Safely crossing an intersection requires that drivers actively control their approach to the intersection with respect to characteristics of the flow of incoming traffic. To further our understanding of the perceptual-motor processes involved in this demanding manoeuvre, we designed a driving simulator experiment in which 13 participants actively negotiated intersections by passing through a gap in the train of incoming traffic. Task constraints were manipulated by varying the size of the traffic gap and the initial conditions with respect to the time of arrival of the traffic gap at the intersection. Environment constraints were manipulated by varying the intersection geometry through changes in the angle formed by the crossroads. The results revealed that the task constraints systematically gave rise to continuous and gradual adjustments in approach velocity, initiated well before arriving at the intersection. These functionally appropriate adjustments allowed the drivers to safely cross the intersection, generally just slightly ahead of the center of the traffic gap. Notwithstanding the fact that the geometry of the intersection did not affect the spatiotemporal constraints of the crossing task, approach behavior varied systematically over geometries, suggesting that drivers rely on the traffic gap's bearing angle. Overall, the pattern of results is indicative of a continuous coupling between perception and action, analogous to that observed in locomotor interception tasks.

Minimal predicted distance: a common metric for collision avoidance during pairwise interactions between walkers

This study investigated collision avoidance between two walkers by focusing on the conditions that lead to avoidance manoeuvres in locomotor trajectories. Following the hypothesis of a reciprocal interaction, we suggested a mutual variable as a continuous function of the two walkers' states, denoted minimum predicted distance (MPD). This function predicts the risk of collision, and its evolution over time captures the motion adaptations performed by the walkers. By groups of two, 30 walkers were assigned locomotion tasks which lead to potential collisions. Results showed that walkers adapted their motions only when required, i.e., when MPD is too low (<1 m). We concluded that walkers are able (i) to accurately estimate their reciprocal distance at the time the crossing will occur, and (ii) to mutually adapt this distance. Furthermore, the study of MPD evolution showed three successive phases in the avoidance interaction: observation where MPD(t) is constant, reaction where MPD(t) increases to acceptable values by adapting locomotion and regulation where MPD(t) reaches a plateau and slightly decreases. This final phase demonstrates that collision avoidance is actually performed with anticipation. Future work would consist in inspecting individual motion adaptations and relating them with the variations of MPD.

Active gaze control improves optic flow-based segmentation and steering

An observer traversing an environment actively relocates gaze to fixate objects. Evidence suggests that gaze is frequently directed toward the center of an object considered as target but more likely toward the edges of an object that appears as an obstacle. We suggest that this difference in gaze might be motivated by specific patterns of optic flow that are generated by either fixating the center or edge of an object. To support our suggestion we derive an analytical model that shows: Tangentially fixating the outer surface of an obstacle leads to strong flow discontinuities that can be used for flow-based segmentation. Fixation of the target center while gaze and heading are locked without head-, body-, or eye-rotations gives rise to a symmetric expansion flow with its center at the point being approached, which facilitates steering toward a target. We conclude that gaze control incorporates ecological constraints to improve the robustness of steering and collision avoidance by actively generating flows appropriate to solve the task.

Coordination dynamics in a socially situated nervous system

Traditional theories of cognitive science have typically accounted for the organization of human behavior by detailing requisite computational/representational functions and identifying neurological mechanisms that might perform these functions. Put simply, such approaches hold that neural activity causes behavior. This same general framework has been extended to accounts of human social behavior via concepts such as “common-coding” and “co-representation” and much recent neurological research has been devoted to brain structures that might execute these social-cognitive functions. Although these neural processes are unquestionably involved in the organization and control of human social interactions, there is good reason to question whether they should be accorded explanatory primacy. Alternatively, we propose that a full appreciation of the role of neural processes in social interactions requires appropriately situating them in their context of embodied-embedded constraints. To this end, we introduce concepts from dynamical systems theory and review research demonstrating that the organization of human behavior, including social behavior, can be accounted for in terms of self-organizing processes and lawful dynamics of animal-environment systems. Ultimately, we hope that these alternative concepts can complement the recent advances in cognitive neuroscience and thereby provide opportunities to develop a complete and coherent account of human social interaction.

Do walkers follow their heads? Investigating the role of head rotation in locomotor control

Eye and head rotations are normally correlated with changes in walking direction; however, it is unknown whether they play a causal role in the control of steering. The objective of the present study was to answer two questions about the role of head rotations in steering control when walking to a goal. First, are head rotations sufficient to elicit a change in walking direction? Second, are head rotations necessary to initiate a change in walking direction or guide steering to a goal? To answer these questions, participants either walked toward a goal located 7 m away or were cued to steer to the left or right by 37°. On a subset of trials, participants were either cued to voluntarily turn their heads to the left or right, or they underwent an involuntary head perturbation via a head-mounted air jet. The results showed that large voluntary head turns (35°) yielded slight path deviations (1°-2°) in the same or opposite direction as the head turn, depending on conditions, which have alternative explanations. Involuntary head rotations did not elicit path deviations despite comparable head rotation magnitudes. In addition, the walking trajectory when turning toward an eccentric goal was the same regardless of head orientation. Steering can thus be decoupled from head rotation during walking. We conclude that head rotations are neither a sufficient nor a necessary component of steering control, because they do not induce a turn and they are not required to initiate a turn or to guide the locomotor trajectory to a goal.

The influence of ground contact and visible horizon on perception of distance and size under severely degraded vision

For low vision navigation, misperceiving the locations of hazards can have serious consequences. Potential sources of such misperceptions are hazards that are not visually associated with the ground plane, thus, depriving the viewer of important perspective cues for egocentric distance. In Experiment 1, we assessed absolute distance and size judgments to targets on stands under degraded vision conditions. Normally sighted observers wore blur goggles that severely reduced acuity and contrast, and viewed targets placed on either detectable or undetectable stands. Participants in the detectable stand condition demonstrated accurate distance judgments, whereas participants in the undetectable stand condition overestimated target distances. Similarly, the perceived size of targets in the undetectable stand condition was judged to be significantly larger than in the detectable stand condition, suggesting a perceptual coupling of size and distance in conditions of degraded vision. In Experiment 2, we investigated size and implied distance perception of targets elevated above a visible horizon for individuals in an induced state of degraded vision. When participants' size judgments are inserted into the size-distance invariance hypothesis (SDIH) formula, distance to above-horizon objects increased compared to those below the horizon. Together, our results emphasize the importance of salient visible ground-contact information for accurate distance perception. The absence of this ground-contact information could be the source of perceptual errors leading to potential hazards for low vision individuals with severely degraded acuity and contrast sensitivity.

Parietal area VIP causally influences heading perception during pursuit eye movements

The ventral intraparietal area (VIP) of the macaque monkey brain is a multimodal area with visual, vestibular, somatosensory, and eye movement-related responses. The visual responses are strongly directional, and VIP neurons respond well to complex optic flow patterns similar to those found during self-motion. To test the hypothesis that visual responses in VIP directly contribute to the perception of self-motion direction, we used electrical microstimulation to perturb activity in VIP while animals performed a two-alternative heading discrimination task. Microstimulation systematically biased monkeys' choices in a direction consistent with neuronal preferences at the stimulation site, and these effects were larger while the animal was making smooth pursuit eye movements. From these results, we conclude that VIP is causally involved in the perception of self-motion from visual cues and that this involvement is gated by ongoing motor behavior.

Monkey steering responses reveal rapid visual-motor feedback

The neural mechanisms underlying primate locomotion are largely unknown. While behavioral and theoretical work has provided a number of ideas of how navigation is controlled, progress will require direct physiolgical tests of the underlying mechanisms. In turn, this will require development of appropriate animal models. We trained three monkeys to track a moving visual target in a simple virtual environment, using a joystick to control their direction. The monkeys learned to quickly and accurately turn to the target, and their steering behavior was quite stereotyped and reliable. Monkeys typically responded to abrupt steps of target direction with a biphasic steering movement, exhibiting modest but transient overshoot. Response latencies averaged approximately 300 ms, and monkeys were typically back on target after about 1 s. We also exploited the variability of responses about the mean to explore the time-course of correlation between target direction and steering response. This analysis revealed a broad peak of correlation spanning approximately 400 ms in the recent past, during which steering errors provoke a compensatory response. This suggests a continuous, visual-motor loop controls steering behavior, even during the epoch surrounding transient inputs. Many results from the human literature also suggest that steering is controlled by such a closed loop. The similarity of our results to those in humans suggests the monkey is a very good animal model for human visually guided steering.

Using vision to control locomotion: looking where you want to go

Looking at the inside edge of the road when steering a bend seems to be a well-established strategy linked to using a feature called the tangent point. An alternative proposal suggests that the gaze patterns observed when steering result from looking at the points in the world through which one wishes to pass. In this explanation fixation on or near the tangent point results from trying to take a trajectory that cuts the corner. To test these accounts, we recorded gaze and steering when taking different paths along curved roadways. Participants could gauge and maintain their lateral distance, but crucially, gaze was predominantly directed to the region proximal to the desired path rather than toward the tangent point per se. These results show that successful control of high-speed locomotion requires fixations in the direction you want to steer rather than using a single road feature like the tangent point.

Using ambulatory virtual environments for the assessment of functional gait impairment: a proof-of-concept study

This study aimed to demonstrate the sensitivity of virtual reality (VR)/motion tracking to detect global functional gait impairment resulting from an emulated knee disability as a prelude to describing mobility changes following lower limb injury/treatment. Participants walked in a figure-8 around two virtual posts placed 6m apart while viewing the computer-generated environment in a helmet-mounted display. Three-dimensional position and orientation of the participant's head were tracked and used to update the virtual scenes, measure walking path and speed, and control task parameters with real-time feedback. Participants walked with/without an emulated lower extremity disability (splint preventing normal knee flexion). Participants performed the task at self-selected Natural (NAT) speed providing a baseline measure of their turning speed and area. Turning speed and area were then in turn maintained fixed (controlled speed, CS; controlled path, CP) while the other variable was measured as a gait impairment indicator. Different adaptive strategies were used to cope with the emulated deficit during the NAT scenario: maintaining turning speed while altering path geometry; decreasing turning speed while maintaining path geometry; and combining the previous two strategies. This resulted, on average, in decreased turning speeds and increased turning areas. The CS and CP manipulations respectively generated even greater turning areas and more consistent speed decreases. The three subtests acted as intertwined filters enabling the detection of functional gait impairment in all subjects regardless of their adaptive strategies. This proof-of-concept study demonstrated how VR/motion tracking technology can be used to detect and quantitatively characterize global functional mobility impairment.

The responses of VIP neurons are sufficiently sensitive to support heading judgments

The ventral intraparietal area (VIP) of the macaque monkey is thought to be involved in judging heading direction based on optic flow. We recorded neuronal discharges in VIP while monkeys were performing a two-alternative, forced-choice heading discrimination task to relate quantitatively the activity of VIP neurons to monkeys' perceptual choices. Most VIP neurons were responsive to simulated heading stimuli and were tuned such that their responses changed across a range of forward trajectories. Using receiver operating characteristic (ROC) analysis, we found that most VIP neurons were less sensitive to small heading changes than was the monkey, although a minority of neurons were equally sensitive. Pursuit eye movements modestly yet significantly increased both neuronal and behavioral thresholds by approximately the same amount. Our results support the view that VIP activity is involved in self-motion judgments.

Towards a new ecological conception of perceptual information: lessons from a developmental systems perspective

Over the last decades or so, empirical studies of perception, action, learning, and development have revealed that participants vary in what variable they detect and often rely on nonspecifying variables. This casts doubt on the Gibsonian conception of information as specification. It is argued that a recent ecological conception of information has solved important problems, but insufficiently explains what determines the object of perception. Drawing on recent work on developmental systems, we sketch the outlines of an alternative conception of perceptual information. It is argued that perceptual information does not reside in the ambient arrays; rather, perceptual information is a relational property of patterns in the array and perceptual processes. What a pattern in the ambient flow informs about depends on the perceiver who uses it. We explore the implications of this alternative conception of information for the ecological approach to perception and action.

The development of locomotor planning for end-state comfort

Walking through real-world environments involves using perceptual information to make complex choices between alternative routes, and this ability must develop through childhood. We examined performance and its development in one such situation. We used a novel 'river-crossing' paradigm analogous to manual 'end-state comfort' planning tasks, where an uncomfortable manoeuvre at the start of a movement is traded off for comfort at its end. Adults showed locomotor end-state comfort planning, adjusting feet at the start of a route in order to gain comfort at its end (crossing a manageable gap between two stepping stones). 3-6-year-olds also made this trade-off, but to a lesser degree than adults. The results suggest that end-state comfort is an important determiner of locomotor behaviour. Furthermore, they show that children as young as 3 years can use detailed visual information to form sophisticated locomotor plans.

On the open-loop and feedback processes that underlie the formation of trajectories during visual and nonvisual locomotion in humans

We investigated the nature of the control mechanisms at work during goal-oriented locomotion. In particular, we tested the effects of vision, locomotor speed, and the presence of via points on the geometric and kinematic properties of locomotor trajectories. We first observed that the average trajectories recorded in visual and nonvisual locomotion were highly comparable, suggesting the existence of vision-independent processes underlying the formation of locomotor trajectories. Then by analyzing and comparing the variability around the average trajectories across different experimental conditions, we were able to demonstrate the existence of on-line feedback control in both visual and nonvisual locomotion and to clarify the relations between visual and nonvisual control strategies. Based on these insights, we designed a model in which maximum-smoothness and optimal feedback control principles account, respectively, for the open-loop and feedback processes. Taken together, the experimental and modeling findings provide a novel understanding of the nature of the motor, sensory, and "navigational" processes underlying goal-oriented locomotion.

Naturalizing Perception

We believe that one of the most important aspects of Gibson's ecological psychology is his attempted naturalization of perception, that is, his attempt to place perception in the context of evolutionary theory. However, the dominant neo-Gibsonian approach to perception has been criticized for being inconsistent with evolutionary theory. We argue that a central tenet of this approach indeed runs counter to evolutionary considerations. Based on an evolutionary analysis of the use of information, we sketch an alternative development of Gibson's pioneering ideas. A truly naturalistic theory of perception, we argue, should recognize both suboptimalities in perception and variation among the members of a population in what informational variables are used. Like other variable organismal features, the use of information is a function of multiple factors. We will compare this naturalistic ecological approach with both Gibson's own perspective and more recent frameworks.

Maneuvers during legged locomotion

Maneuverability is essential for locomotion. For animals in the environment, maneuverability is directly related to survival. For humans, maneuvers such as turning are associated with increased risk for injury, either directly through tissue loading or indirectly through destabilization. Consequently, understanding the mechanics and motor control of maneuverability is a critical part of locomotion research. We briefly review the literature on maneuvering during locomotion with a focus on turning in bipeds. Walking turns can use one of several different strategies. Anticipation can be important to adjust kinematics and dynamics for smooth and stable maneuvers. During running, turns may be substantially constrained by the requirement for body orientation to match movement direction at the end of a turn. A simple mathematical model based on the requirement for rotation to match direction can describe leg forces used by bipeds (humans and ostriches). During running turns, both humans and ostriches control body rotation by generating fore-aft forces. However, whereas humans must generate large braking forces to prevent body over-rotation, ostriches do not. For ostriches, generating the lateral forces necessary to change movement direction results in appropriate body rotation. Although ostriches required smaller braking forces due in part to increased rotational inertia relative to body mass, other movement parameters also played a role. Turning performance resulted from the coordinated behavior of an integrated biomechanical system. Results from preliminary experiments on horizontal-plane stabilization support the hypothesis that controlling body rotation is an important aspect of stable maneuvers. In humans, body orientation relative to movement direction is rapidly stabilized during running turns within the minimum of two steps theoretically required to complete analogous maneuvers. During straight running and cutting turns, humans exhibit spring-mass behavior in the horizontal plane. Changes in the horizontal projection of leg length were linearly related to changes in horizontal-plane leg forces. Consequently, the passive dynamic stabilization associated with spring-mass behavior may contribute to stability during maneuvers in bipeds. Understanding the mechanics of maneuverability will be important for understanding the motor control of maneuvers and also potentially be useful for understanding stability.

Cortical dynamics of navigation and steering in natural scenes: Motion-based object segmentation, heading, and obstacle avoidance

Visually guided navigation through a cluttered natural scene is a challenging problem that animals and humans accomplish with ease. The ViSTARS neural model proposes how primates use motion information to segment objects and determine heading for purposes of goal approach and obstacle avoidance in response to video inputs from real and virtual environments. The model produces trajectories similar to those of human navigators. It does so by predicting how computationally complementary processes in cortical areas MT(-)/MSTv and MT(+)/MSTd compute object motion for tracking and self-motion for navigation, respectively. The model's retina responds to transients in the input stream. Model V1 generates a local speed and direction estimate. This local motion estimate is ambiguous due to the neural aperture problem. Model MT(+) interacts with MSTd via an attentive feedback loop to compute accurate heading estimates in MSTd that quantitatively simulate properties of human heading estimation data. Model MT(-) interacts with MSTv via an attentive feedback loop to compute accurate estimates of speed, direction and position of moving objects. This object information is combined with heading information to produce steering decisions wherein goals behave like attractors and obstacles behave like repellers. These steering decisions lead to navigational trajectories that closely match human performance.

Limitations of feedforward control in multiple-phase steering movements

When attempting to perform bi-phasic steering movements (such as a lane change) in the absence of visual and inertial feedback, drivers produce a systematic heading error in the direction of the lane change (Wallis et al., Curr Biol 12(4):295-299, 2002; J Exp Psychol Hum Percept Perform 33(55):1127-1144, 2007). Theories of steering control which employ exclusively open-loop control mechanisms cannot accommodate this finding. In this article we show that a similar steering error occurs with obstacle avoidance, and offer compelling evidence that it stems from a seemingly general failure of human operators to correctly internalise the dynamics of the steering wheel. With respect to lateral position, the steering wheel is an acceleration control device, but we present data indicating that drivers treat it as a rate control device. Previous findings from Wallis et al. can be explained the same way. Since an open-loop control mechanism will never succeed when the dynamics of the controller are internalised improperly, we go on to conclude that regular, appropriately timed sensory feedback-predominantly from vision-is necessary for regulating heading, even during well-practiced, everyday manoeuvres such as lane changing and obstacle avoidance.

Behaviour and Gaze Analyses During a Goal-Directed Locomotor Task

The objectives of the current study were: (a) to determine whether perception–action coupling controlled behaviours when walking through moving doors and (b) to determine how vision contributed to this behaviour. Participants ( N = 6) walked along a 7-m path toward two motor-driven doors, which moved at rates ranging between 20 and 40 cm/s. Each door was independently driven such that both moved at the same velocity (symmetrical) or at different velocities (asymmetrical). The results showed that in both door movement conditions the participants controlled their approach velocity by slowing down prior to crossing the doors. The decrease in walking velocity produced greater velocity variability in the final stages prior to crossing the doors and high success rates. The results from the gaze behaviours showed that fixation durations were significantly longer when the doors moved asymmetrically, suggesting that the visual information from this unpredictable environment took longer to process. However, the fixation patterns were similar between the two door movement conditions. Regardless of the door movement condition, the participants spent about 60% of each trial fixating environmental objects (i.e., left door, right door, or aperture). The majority of fixations were directed towards one of the doors at the beginning of the trial and then shifted towards the aperture in the final phase. The participants were using perception–action coupling to control their behaviours in the final phase in order to steer locomotion through the aperture.

Locomotor avoidance behaviours during a visually guided task involving an approaching object

Collision avoidance behaviours in situations where a collision may occur and one's planned movement is restricted, reveals that one's response is not as simple as a visual input producing some motor output. In this study, the participants (N=6) walked along a 9.5m path towards an air-filled human doll (180 degrees from their travel path) that would approach them on some trials. A spatial constraint (i.e. doorframe) was placed along the path and the participants had to determine if they could pass through the constraint prior to avoiding a collision or not. The constraint was set-up so that it was either at the theoretical collision point or 1.5m before or after the theoretical collision point. This study aimed to determine: (1) how the presence of a spatial constraint affects one's ability to perceive when to avoid a collision with an approaching object; (2) if the individuals use action parameters (i.e. velocity modifications, change in heading, etc.) in a consistent manner independent of the spatial constraint location and object's approach velocity; (3) if a consistent safety zone exists independent of the object's approach velocity. The results showed that the placement of the spatial constraint, but not the velocity of the object had a significant effect on the initiation of a change in heading. Participants used two-stage avoidance behaviour; change heading and then adjust walking velocity. The initial avoidance behaviour was initiated when the object was at a constant distance away (i.e. 3.73 m). Overall, it appears as though collision avoidance with approaching objects has cognitive as well as perceptual influences.

Testing the role of expansion in the prospective control of locomotion

The constant bearing angle (CBA) strategy is a prospective strategy that permits the interception of moving objects. The purpose of the present study is to test this strategy. Participants were asked to walk through a virtual environment and to change, if necessary, their walking speed so as to intercept approaching targets. The targets followed either a rectilinear or a curvilinear trajectory and target size was manipulated both within trials (target size was gradually changed during the trial in order to bias expansion) and between trials (targets of different sizes were used). The curvature manipulation had a large effect on the kinematics of walking, which is in agreement with the CBA strategy. The target size manipulations also affected the kinematics of walking. Although these effects of target size are not predicted by the CBA strategy, quantitative comparisons of observed kinematics and the kinematics predicted by the CBA strategy showed good fits. Furthermore, predictions based on the CBA strategy were deemed superior to predictions based on a required velocity (V (REQ)) model. The role of target size and expansion in the prospective control of walking is discussed.

Flying Fast and Low Among Obstacles: Methodology and Experiments

Safe autonomous flight is essential for widespread acceptance of aircraft that must fly close to the ground. We have developed a method of collision avoidance that can be used in three dimensions in much the same way as autonomous ground vehicles that navigate over unexplored terrain. Safe navigation is accomplished by a combination of online environmental sensing, path planning and collision avoidance. Here we outline our methodology and report results with an autonomous helicopter that operates at low elevations in uncharted environments, some of which are densely populated with obstacles such as buildings, trees and wires. We have recently completed over 700 successful runs in which the helicopter traveled between coarsely specified waypoints separated by hundreds of meters, at speeds of up to 10 m s—1 at elevations of 5—11 m above ground level. The helicopter safely avoids large objects such as buildings and trees but also wires as thin as 6 mm. We believe this represents the first time an air vehicle has traveled this fast so close to obstacles. The collision avoidance method learns to avoid obstacles by observing the performance of a human operator.

Strategies used to walk through a moving aperture

The objectives of the study were to determine what strategy (pursuit or interception) individuals used to pass through an oscillating target and to determine if individuals walked towards where they were looking. Kinematic and gaze behaviour data was collected from seven healthy female participants as they started at one of five different starting positions and walked 7 m towards an oscillating target. The target was a two-dimensional 70 cm aperture made by two-76 cm wide doors and oscillated between two end posts that were 300 cm apart. In order to quantify the objectives, target-heading angles [Fajen BR, Warren WH. Behavioral dynamics of steering, obstacle avoidance, and route selection. J Exp Psychol Hum Percept Perform 2003;29(2):343-62; Fajen BR, Warren WH. Visual guidance of intercepting a moving target on foot. Perception 2004;33:689-715] were calculated. Results showed that the participants used neither an interception nor a pursuit strategy to successfully pass through the moving aperture. The participants steered towards the middle of the pathway prior to passing through the middle of the aperture. A cross correlation between the horizontal gaze locations and the medial/lateral (M/L) location of the participants' center of mass (COM) was performed. The results from the cross correlation show that during the final 2s prior to crossing the aperture, the participants walked where they were looking. The findings from this study suggest that individuals simplify a task by decreasing the perceptual load until the final stages. In this way the final stages of this task were visually driven.

Characteristics of personal space during obstacle circumvention in physical and virtual environments

It is not known how the flexible protective zone maintained around oneself during locomotion (personal space or PS; see [Gérin-Lajoie M, Richards CL, McFadyen BJ. The negotiation of stationary and moving obstructions during walking: anticipatory locomotor adaptations and preservation of personal space. Motor Control 2005;9:242-69]) is modulated with walking speed, whether both sides of the PS are symmetrical, and whether the circumvention of physical and virtual obstructions elicit the same use of such PS. Personal space was measured in ten adults as they circumvented a cylindrical obstacle that was stationary within their path. Both left and right passes were performed at natural self-selected, slow and fast walking speeds. The same circumvention task was also performed at natural speeds in an immersive virtual environment (VE) replicating the same obstruction scenario. The shape and size of PS were maintained across walking speeds, and a smaller PS was generally observed on the dominant side. The general shape and lateral bias of the PS were preserved in the VE while its size was slightly increased. The systematic behavior across walking speeds and types of environment and the lateral bias suggest that PS is used to control navigation. This study deepens our understanding of normal adaptive walking behavior and has implications for the development of better tools for the assessment and retraining of locomotor capacity in different populations, from people with walking deficits to elite athletes. Since the PS behavior was shown to be robust in the VE used for this study, the virtual reality technology is proposed as a promising platform for the development of such assessment and retraining applications.