76
|
Gomez-Gonzalez S, Neumann G, Scholkopf B, Peters J. Adaptation and Robust Learning of Probabilistic Movement Primitives. IEEE T ROBOT 2020. [DOI: 10.1109/tro.2019.2937010] [Citation(s) in RCA: 19] [Impact Index Per Article: 4.8] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/11/2022]
|
77
|
Veiga F, Edin B, Peters J. Grip Stabilization through Independent Finger Tactile Feedback Control. SENSORS 2020; 20:s20061748. [PMID: 32245193 PMCID: PMC7146159 DOI: 10.3390/s20061748] [Citation(s) in RCA: 11] [Impact Index Per Article: 2.8] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 01/31/2020] [Revised: 03/17/2020] [Accepted: 03/19/2020] [Indexed: 11/16/2022]
Abstract
Grip force control during robotic in-hand manipulation is usually modeled as a monolithic task, where complex controllers consider the placement of all fingers and the contact states between each finger and the gripped object in order to compute the necessary forces to be applied by each finger. Such approaches normally rely on object and contact models and do not generalize well to novel manipulation tasks. Here, we propose a modular grip stabilization method based on a proposition that explains how humans achieve grasp stability. In this biomimetic approach, independent tactile grip stabilization controllers ensure that slip does not occur locally at the engaged robot fingers. Local slip is predicted from the tactile signals of each fingertip sensor i.e., BioTac and BioTac SP by Syntouch. We show that stable grasps emerge without any form of central communication when such independent controllers are engaged in the control of multi-digit robotic hands. The resulting grasps are resistant to external perturbations while ensuring stable grips on a wide variety of objects.
Collapse
|
78
|
Stein A, Peers E, Hattersley J, Harris K, Feehally J, Walls J, Brown C, Nichols N, Mistry C, Ashman R, Raftery M, Peters J, Gokal R. Clinical Experience with Icodextrin in Continuous Ambulatory Peritoneal Dialysis Patients. Perit Dial Int 2020. [DOI: 10.1177/089686089401402s08] [Citation(s) in RCA: 14] [Impact Index Per Article: 3.5] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/17/2022] Open
Abstract
Objectives To review the clinical experience in the United Kingdom with icodextrin (Ic). Design A retrospective multicenter study. Patients (1) the MICAS 1 patients who received Ic and elected to continue using it (called MICAS 2 patients), and (2) patients started on Ic on a named-patient basis (called compassionate use patients). Nearly all this latter group had ultrafiltration (UF) failure and were approaching hemodialysis (HC). Interventions The institution of one Ic bag, as the overnight exchange, usually in addition to three glucose bags. Results (1) MICAS 2: Of the 67 patients who received Ic and completed MICAS 1, 48 (72%) patients from eight units in the United Kingdom wished to continue Ic. Forty two percent of MICAS 2 patients were well on Ic after approximately 30 months of study (including the 6 months on MICAS 1). The main reasons for withdrawal were transplantation (21 %), death (17%), or transfer to HC for reasons other than UF failure (10%). There were no significant changes in laboratory data, including hematology, simple biochemistry, bone biochemistry, liver function tests, serum osmolality, and maltose levels. The adverse events reported were incidental to Ic. (2) Compassionate use: 30/63 (48%) patients remain well on Ic after 10 months of study. Eight patients (12%) went on to experience UF failure (again) after 13 months on Ic, requiring transfer to HC. The other main outcomes were death (11 %), transplantation (11 %), or transfer to HC for reasons unrelated to Ic (16%), usually intractable peritonitis. The serum sodium concentration decreased from 136.6±3.8 to 134.4±4.2 mmol/L (p < 0.05). There were no other significant changes in hematology or biochemistry measurements. Conclusion In MICAS 2 Ic probably maintained its effects, since there were no withdrawals due to UF failure. The compassionate use program has shown that a single Ic exchange is useful in UF failure, being able to prevent or delay transfer to hemodialysis. In both studies the safety and efficacy of Ic has been demonstrated in a large group of patients. A total of 192 patient-years of experience (including MICAS 1) has been accumulated to date. Properly controlled studies are needed to determine whether Ic, in its present form, should remain a “special needs” product (for UF failure) or whether it could replace glucose as the treatment of choice for CAPC patients.
Collapse
|
79
|
Gokal R, Mistry CD, Peers E, Brown C, Smith S, Edwards D, Junor B, Gordon A, McMillan M, Robertson M, Michael J, McKain J, Raftery M, Peters J, Clutterbuck E, Clemenger M, Walls J, Orton C, Goodship T, Grieves J, Dharmasena D, Hourhane G, Howarth D, Boyes R, Clisby L, Beran Y. A United Kingdom Multicenter Study of Icodextrin in Continuous Ambulatory Peritoneal Dialysis. Perit Dial Int 2020. [DOI: 10.1177/089686089401402s03] [Citation(s) in RCA: 21] [Impact Index Per Article: 5.3] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/16/2022] Open
Abstract
While glucose remains the only osmotic agent used universally for peritoneal dialysis, its various shortcomings for the long dwell equilibration continuous ambulatory peritoneal dialysis (CAPD) has led to a search for alternative agents. The large molecular weight group has been of interest, because these agents theoretically would lead to greater ultrafiltration and a better metabolic profile. Mostsubstances (dextrans, charged macromolecules) have been found unsuitable for reasons of insolubility, allergenicity, and peritoneal toxicity. Short-chain polypeptides have been studied in humans, but the experience is limited, and there is the potential for allergenicity with long-term use. The only large molecular weight agent that has been studied in some detail but hitherto in one center only and in a limited number of patients is glucose polymer (generic name, icodextrin). Because of the promise shown by these initial studies, a randomized controlled multicenter investigation of icodextrin in CAPD (MIDAS Study Group) was undertaken to evaluate the long-term safety and efficacy by comparing daily overnight (8 12 hours) use of a slightly hypo-osmolar solution (282 mOsm/ kg) with 1.36% (346 mOsm/kg) and 3.86% (484 mOsm/kg) glucose exchanges. Over a 6-month period 209 patients from 11 centers in the United Kingdom were randomized, with 106 allocated to receive icodextrin (study group) and 103 to remain on glucose (control group). One hundred and thirty-eight patients completed the 6-month study (71 control, 67 study). The mean net ultrafiltration overnight with icodextrin was 3.5 times greater than 1.36% at 8 hours and 5.5 times greater at 12 hours (p<0.0001), but no different from that of 3.86% glucose at 8 and 12 hours (although for the latter dwell the net mean ultrafiltration volume was greater by about 140 mL). Biochemical profiles were no different except for a small fall in serum sodium and chloride in the icodextrin group. The mean serum maltose rose to a steady-state level of 1.2 g/L within 2 weeks and remained stable. The mean carbohydrate absorbed for icodextrin (29±5 g) was lower than with 3.86% glucose (62±5 g). The use of icodextrin did not increase the incidence of peritonitis, nor did it alter its outcome, affect uptake of icodextrin from the peritoneum, alter serum osmolality or sodium levels. There were no adverse effects associated with the use of icodextrin, and the overall CAPD-related symptom score was significantly better for icodextrin than control subjects. This study and subsequent extensive use and clinical experience has demonstrated that the daily use of an iso-osmolar icodextrin solution is generally well tolerated, effective, and could replace the overnight use of hyperosmotic glucose solution. Its use was of equal efficacy in peritonitis and in diabetic patients. The elevated levels of maltose did not appear to have any clinical side effects.
Collapse
|
80
|
Gokal R, Mistry CD, Peers EM, Brown C, Smith S, Edwards D, Junor B, Gordon A, McMillan M, Robertson M, Michael J, McKain J, Raftery M, Peters J, Clutterbuck E, Clemenger M, Walls J, Orton C, Goodship T, Grieves J, Olubodun J, Jackson F, Dharmasena D, Hourahane G, Howarth D, Boyes R, Clisby L, Beran Y. Peritonitis Occurrence in a Multicenter Study of Icodextrin and Glucose in Capd. Perit Dial Int 2020. [DOI: 10.1177/089686089501506s07] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.3] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/15/2022] Open
Abstract
Objective To compare peritonitis occurrence and outcome in a large U.K. study Multicentre Investigation of Icodextrin in Ambulatory Dialysis (MIDAS). Design Prospective, randomized, controlled 6-month comparison of icodextrin with glucose for the long dwell in continuous ambulatory peritoneal dialysis (CAPD) patients. Setting Eleven CAPD units in U.K. teaching hospitals. Patients A total of 209 patients established on CAPD for at least 3 months (103 control, 106 icodextrin). Twentythree control (C) and 22 icodextrin (I) patients experienced peritonitis during the study. Intervention Patients who had peritonitis remained on treatment (unless CAPD was withdrawn, temporarily or permanently). Main Outcome Measures The main outcome measures were the rate of peritonitis and duration of CAPD treatment prestudy; the rate of peritonitis episodes and their outcome during study; the effect of peritonitis on laboratory variables, serum icodextrin metabolites, and ultrafiltration efficacy. Results Prestudy: Nine (39%) of C but 14 (64%) of I patients had suffered previous peritonitis episode(s), with overall rates of 0.58 and 0.78 episodes per patientyear, respectively. During study There were 31 C episodes and 35 I episodes, with overall rates of 0.76 and 0.93 per patientyear, respectively. The increase in the C and I groups was 31% and 19%, respectively. Serum osmolality and sodium levels were unaffected by peritonitis, and there was no increase in serum icodextrin metabolites during peritonitis. Overnight ultrafiltration volume during peritonitis (mean±SD) declined slightly from 218±354 mL to 185±299 mL (NS) in the control group, but increased in the icodextrin group from 570±146 mL to 723±218 mL (p < 0.01). Conclusions Using icodextrin for the long dwell in CAPD does not increase the rate of peritonitis, nor does it alter the outcome of peritonitis. Peritonitis does not affect uptake of icodextrin from the peritoneum.
Collapse
|
81
|
Koert D, Trick S, Ewerton M, Lutter M, Peters J. Incremental Learning of an Open-Ended Collaborative Skill Library. INT J HUM ROBOT 2020. [DOI: 10.1142/s0219843620500012] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/18/2022]
Abstract
Intelligent assistive robots can potentially contribute to maintaining an elderly person’s independence by supporting everyday life activities. However, the number of different and personalized activities to be supported renders pre-programming of all respective robot behaviors prohibitively difficult. Instead, to cope with a continuous and potentially open-ended stream of cooperative tasks, new collaborative robot behaviors need to be continuously learned and updated from demonstrations. To this end, we introduce an online learning method to incrementally build a cooperative skill library of probabilistic interaction primitives. The resulting model chooses a corresponding robot response to a human movement where the human intention is extracted from previously demonstrated movements. While existing batch learning methods for movement primitives usually learn such skill libraries only once for a pre-defined number of different skills, our approach enables extending the skill library in an open-ended and online fashion from new incoming demonstrations. The proposed approach is evaluated on a low-dimensional benchmark task and in a collaborative scenario with a 7DoF robot, where we also investigate the generalization of learned skills between different subjects.
Collapse
|
82
|
Thielmann M, Vaghiri S, Dirkmann D, Neuhäuser M, Peters J, Heusch G, Jakob H, Ruhparwar A, Kamler M, Kleinbongard P. Remote Ischemic Preconditioning in Elective Cardiac Surgery: Long-Term Overall Survival Benefit in a Single-Center Randomized Double-Blinded Controlled Trial. Thorac Cardiovasc Surg 2020. [DOI: 10.1055/s-0040-1705456] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 10/24/2022]
|
83
|
Gruber S, Vorhauer N, Schulz M, Hilmer M, Peters J, Tsotsas E, Foerst P. Estimation of the local sublimation front velocities from neutron radiography and tomography of particulate matter. Chem Eng Sci 2020. [DOI: 10.1016/j.ces.2019.115268] [Citation(s) in RCA: 5] [Impact Index Per Article: 1.3] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 10/25/2022]
|
84
|
Vinogradska J, Bischoff B, Achterhold J, Koller T, Peters J. Numerical Quadrature for Probabilistic Policy Search. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 2020; 42:164-175. [PMID: 30403621 DOI: 10.1109/tpami.2018.2879335] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 06/08/2023]
Abstract
Learning control policies has become an appealing alternative to the derivation of control laws based on classic control theory. Model-based approaches have proven an outstanding data efficiency, especially when combined with probabilistic models to eliminate model bias. However, a major difficulty for these methods is that multi-step-ahead predictions typically become intractable for larger planning horizons and can only poorly be approximated. In this paper, we propose the use of numerical quadrature to overcome this drawback and provide significantly more accurate multi-step-ahead predictions. As a result, our approach increases data efficiency and enhances the quality of learned policies. Furthermore, policy learning is not restricted to optimizing locally around one trajectory, as numerical quadrature provides a principled approach to extend optimization to all trajectories starting in a specified starting state region. Thus, manual effort, such as choosing informative starting points for simultaneous policy optimization, is significantly decreased. Furthermore, learning is highly robust to the choice of initial policy and, thus, interaction time with the system is minimized. Empirical evaluations on simulated benchmark problems show the efficiency of the proposed approach and support our theoretical results.
Collapse
|
85
|
Yang J, Duursma RA, De Kauwe MG, Kumarathunge D, Jiang M, Mahmud K, Gimeno TE, Crous KY, Ellsworth DS, Peters J, Choat B, Eamus D, Medlyn BE. Incorporating non-stomatal limitation improves the performance of leaf and canopy models at high vapour pressure deficit. TREE PHYSIOLOGY 2019; 39:1961-1974. [PMID: 31631220 DOI: 10.1093/treephys/tpz103] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.2] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Subscribe] [Scholar Register] [Received: 05/08/2019] [Revised: 08/02/2019] [Accepted: 09/12/2019] [Indexed: 06/10/2023]
Abstract
Vapour pressure deficit (D) is projected to increase in the future as temperature rises. In response to increased D, stomatal conductance (gs) and photosynthesis (A) are reduced, which may result in significant reductions in terrestrial carbon, water and energy fluxes. It is thus important for gas exchange models to capture the observed responses of gs and A with increasing D. We tested a series of coupled A-gs models against leaf gas exchange measurements from the Cumberland Plain Woodland (Australia), where D regularly exceeds 2 kPa and can reach 8 kPa in summer. Two commonly used A-gs models were not able to capture the observed decrease in A and gs with increasing D at the leaf scale. To explain this decrease in A and gs, two alternative hypotheses were tested: hydraulic limitation (i.e., plants reduce gs and/or A due to insufficient water supply) and non-stomatal limitation (i.e., downregulation of photosynthetic capacity). We found that the model that incorporated a non-stomatal limitation captured the observations with high fidelity and required the fewest number of parameters. Whilst the model incorporating hydraulic limitation captured the observed A and gs, it did so via a physical mechanism that is incorrect. We then incorporated a non-stomatal limitation into the stand model, MAESPA, to examine its impact on canopy transpiration and gross primary production. Accounting for a non-stomatal limitation reduced the predicted transpiration by ~19%, improving the correspondence with sap flow measurements, and gross primary production by ~14%. Given the projected global increases in D associated with future warming, these findings suggest that models may need to incorporate non-stomatal limitation to accurately simulate A and gs in the future with high D. Further data on non-stomatal limitation at high D should be a priority, in order to determine the generality of our results and develop a widely applicable model.
Collapse
|
86
|
Lioutikov R, Maeda G, Veiga F, Kersting K, Peters J. Learning attribute grammars for movement primitive sequencing. Int J Rob Res 2019. [DOI: 10.1177/0278364919868279] [Citation(s) in RCA: 7] [Impact Index Per Article: 1.4] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/15/2022]
Abstract
Movement primitives are a well studied and widely applied concept in modern robotics. However, composing primitives out of an existing library has shown to be a challenging problem. We propose the use of probabilistic context-free grammars to sequence a series of primitives to generate complex robot policies from a given library of primitives. The rule-based nature of formal grammars allows an intuitive encoding of hierarchically structured tasks. This hierarchical concept strongly connects with the way robot policies can be learned, organized, and re-used. However, the induction of context-free grammars has proven to be a complicated and yet unsolved challenge. We exploit the physical nature of robot movement primitives to restrict and efficiently search the grammar space. The grammar is learned by applying a Markov chain Monte Carlo optimization over the posteriors of the grammars given the observations. The proposal distribution is defined as a mixture over the probabilities of the operators connecting the search space. Moreover, we present an approach for the categorization of probabilistic movement primitives and discuss how the connectibility of two primitives can be determined. These characteristics in combination with restrictions to the operators guarantee continuous sequences while reducing the grammar space. In addition, a set of attributes and conditions is introduced that augments probabilistic context-free grammars in order to solve primitive sequencing tasks with the capability to adapt single primitives within the sequence. The method was validated on tasks that require the generation of complex sequences consisting of simple movement primitives using a seven-degree-of-freedom lightweight robotic arm.
Collapse
|
87
|
Natali F, Dolce C, Peters J, Stelletta C, Demé B, Ollivier J, Leduc G, Cupane A, Barbier EL. Brain lateralization probed by water diffusion at the atomic to micrometric scale. Sci Rep 2019; 9:14694. [PMID: 31604980 PMCID: PMC6789030 DOI: 10.1038/s41598-019-51022-1] [Citation(s) in RCA: 3] [Impact Index Per Article: 0.6] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/17/2019] [Accepted: 09/23/2019] [Indexed: 01/27/2023] Open
Abstract
Combined neutron scattering and diffusion nuclear magnetic resonance experiments have been used to reveal significant interregional asymmetries (lateralization) in bovine brain hemispheres in terms of myelin arrangement and water dynamics at micron to atomic scales. Thicker myelin sheaths were found in the left hemisphere using neutron diffraction. 4.7 T dMRI and quasi-elastic neutron experiments highlighted significant differences in the properties of water dynamics in the two hemispheres. The results were interpreted in terms of hemisphere-dependent cellular composition (number of neurons, cell distribution, etc.) as well as specificity of neurological functions (such as preferential networking).
Collapse
|
88
|
Ahn M, Cantarini M, Frewer P, Hawkins G, Peters J, Howarth P, Ahmed G, Sahota T, Hartmaier R, Li-Sucholeiki X, Oxnard G. P1.01-134 SAVANNAH: Phase II Trial of Osimertinib + Savolitinib in EGFR-Mutant, MET-Driven Advanced NSCLC, Following Prior Osimertinib. J Thorac Oncol 2019. [DOI: 10.1016/j.jtho.2019.08.849] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.4] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/29/2022]
|
89
|
Koert D, Pajarinen J, Schotschneider A, Trick S, Rothkopf C, Peters J. Learning Intention Aware Online Adaptation of Movement Primitives. IEEE Robot Autom Lett 2019. [DOI: 10.1109/lra.2019.2928760] [Citation(s) in RCA: 6] [Impact Index Per Article: 1.2] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/10/2022]
|
90
|
Ewerton M, Arenz O, Maeda G, Koert D, Kolev Z, Takahashi M, Peters J. Learning Trajectory Distributions for Assisted Teleoperation and Path Planning. Front Robot AI 2019; 6:89. [PMID: 33501104 PMCID: PMC7806068 DOI: 10.3389/frobt.2019.00089] [Citation(s) in RCA: 6] [Impact Index Per Article: 1.2] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/31/2019] [Accepted: 09/03/2019] [Indexed: 11/13/2022] Open
Abstract
Several approaches have been proposed to assist humans in co-manipulation and teleoperation tasks given demonstrated trajectories. However, these approaches are not applicable when the demonstrations are suboptimal or when the generalization capabilities of the learned models cannot cope with the changes in the environment. Nevertheless, in real co-manipulation and teleoperation tasks, the original demonstrations will often be suboptimal and a learning system must be able to cope with new situations. This paper presents a reinforcement learning algorithm that can be applied to such problems. The proposed algorithm is initialized with a probability distribution of demonstrated trajectories and is based on the concept of relevance functions. We show in this paper how the relevance of trajectory parameters to optimization objectives is connected with the concept of Pearson correlation. First, we demonstrate the efficacy of our algorithm by addressing the assisted teleoperation of an object in a static virtual environment. Afterward, we extend this algorithm to deal with dynamic environments by utilizing Gaussian Process regression. The full framework is applied to make a point particle and a 7-DoF robot arm autonomously adapt their movements to changes in the environment as well as to assist the teleoperation of a 7-DoF robot arm in a dynamic environment.
Collapse
|
91
|
Malewicz NM, Walstein K, Heine T, Engler A, Bick A, Cox L, Dötsch A, Westendorf AM, Horn PA, Lindemann M, Peters J, Schäfer ST. Early suppression of peripheral mononuclear blood cells in sepsis in response to stimulation with cytomegalovirus, OKT3, and pokeweed mitogen. J Appl Physiol (1985) 2019; 127:1539-1547. [PMID: 31545153 DOI: 10.1152/japplphysiol.00438.2019] [Citation(s) in RCA: 3] [Impact Index Per Article: 0.6] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 12/17/2022] Open
Abstract
Critically ill patients are at risk for sepsis, and immunosuppressive mechanisms may prevail. Whether functional tests are helpful to detect immune alterations is largely unknown. Therefore, we tested the hypotheses that reactivity of peripheral blood mononuclear cells (PBMCs) to secrete interferon-γ (IFNγ) following stimulation in vitro is decreased in patients with early sepsis compared with postoperative patients. IFNγ secretion [enzyme-linked immunospot (ELISpot)] in response to stimulation with cytomegalovirus (CMV), pokeweed mitogen (PWM), muromonab-anti-CD3 (OKT3), and human leukocyte antigen (HLA)-DRA-mRNA expression and serum cytokine concentrations were repeatedly [days 1, 3, 5, and 7 after intensive care unit (ICU) admission] determined in patients with sepsis (n = 7) and patients undergoing major abdominal surgery (radical prostatectomy, cystectomy, n = 10). In a second cohort, HLA-DRA expression was assessed in 80 patients with sepsis, 30 postoperative patients, and 44 healthy volunteers (German clinical trials database no. 00007694). In patients with sepsis, IFNγ secretion (ELISpot) was decreased compared with controls after stimulation with CMV (P = 0.01), OKT3 (P = 0.02), and PWM (P = 0.02 on day 5), whereas unstimulated IFNγ secretion did not differ. HLA-DRA expression was also significantly decreased in patients with sepsis at all time points (P = 0.004) compared with postoperative surgical patients, a finding confirmed in the larger cohort. Reactivity of PBMCs to stimulation with CMV, PWM, and OKT3 as well as HLA-DRA expression was already decreased upon ICU admission in patients with sepsis when compared with postoperative controls, suggesting early depression of acquired immunity. ELISpot assays may help to clinically characterize the time course of immunocompetence in patients with sepsis.NEW & NOTEWORTHY We observed suppression of reactivity to stimulation with cytomegalovirus, muromonab-anti-CD3, and pokeweed mitogen in mononuclear blood cells of patients with early sepsis when compared with postoperative controls. Thus, there is early depression of acquired immunity in sepsis. Enzyme-linked immunospot assays may help to characterize immunocompetence in patients with sepsis.
Collapse
|
92
|
Celemin C, Maeda G, Ruiz-del-Solar J, Peters J, Kober J. Reinforcement learning of motor skills using Policy Search and human corrective advice. Int J Rob Res 2019. [DOI: 10.1177/0278364919871998] [Citation(s) in RCA: 9] [Impact Index Per Article: 1.8] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/16/2022]
Abstract
Robot learning problems are limited by physical constraints, which make learning successful policies for complex motor skills on real systems unfeasible. Some reinforcement learning methods, like Policy Search, offer stable convergence toward locally optimal solutions, whereas interactive machine learning or learning-from-demonstration methods allow fast transfer of human knowledge to the agents. However, most methods require expert demonstrations. In this work, we propose the use of human corrective advice in the actions domain for learning motor trajectories. Additionally, we combine this human feedback with reward functions in a Policy Search learning scheme. The use of both sources of information speeds up the learning process, since the intuitive knowledge of the human teacher can be easily transferred to the agent, while the Policy Search method with the cost/reward function take over for supervising the process and reducing the influence of occasional wrong human corrections. This interactive approach has been validated for learning movement primitives with simulated arms with several degrees of freedom in reaching via-point movements, and also using real robots in such tasks as “writing characters” and the ball-in-a-cup game. Compared with standard reinforcement learning without human advice, the results show that the proposed method not only converges to higher rewards when learning movement primitives, but also that the learning is sped up by a factor of 4–40 times, depending on the task.
Collapse
|
93
|
Natali F, Dolce C, Peters J, Stelletta C, Demé B, Ollivier J, Boehm M, Leduc G, Piazza I, Cupane A, Barbier EL. Anomalous water dynamics in brain: a combined diffusion magnetic resonance imaging and neutron scattering investigation. J R Soc Interface 2019; 16:20190186. [PMID: 31409238 DOI: 10.1098/rsif.2019.0186] [Citation(s) in RCA: 10] [Impact Index Per Article: 2.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 12/31/2022] Open
Abstract
Water diffusion is an optimal tool for investigating the architecture of brain tissue on which modern medical diagnostic imaging techniques rely. However, intrinsic tissue heterogeneity causes systematic deviations from pure free-water diffusion behaviour. To date, numerous theoretical and empirical approaches have been proposed to explain the non-Gaussian profile of this process. The aim of this work is to shed light on the physics piloting water diffusion in brain tissue at the micrometre-to-atomic scale. Combined diffusion magnetic resonance imaging and first pioneering neutron scattering experiments on bovine brain tissue have been performed in order to probe diffusion distances up to macromolecular separation. The coexistence of free-like and confined water populations in brain tissue extracted from a bovine right hemisphere has been revealed at the micrometre and atomic scale. The results are relevant for improving the modelling of the physics driving intra- and extracellular water diffusion in brain, with evident benefit for the diffusion magnetic resonance imaging technique, nowadays widely used to diagnose, at the micrometre scale, brain diseases such as ischemia and tumours.
Collapse
|
94
|
Koc O, Maeda G, Peters J. Optimizing the Execution of Dynamic Robot Movements With Learning Control. IEEE T ROBOT 2019. [DOI: 10.1109/tro.2019.2906558] [Citation(s) in RCA: 7] [Impact Index Per Article: 1.4] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/06/2022]
|
95
|
Belousov B, Peters J. Entropic Regularization of Markov Decision Processes. ENTROPY 2019; 21:e21070674. [PMID: 33267388 PMCID: PMC7515171 DOI: 10.3390/e21070674] [Citation(s) in RCA: 4] [Impact Index Per Article: 0.8] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 06/14/2019] [Revised: 07/06/2019] [Accepted: 07/08/2019] [Indexed: 11/16/2022]
Abstract
An optimal feedback controller for a given Markov decision process (MDP) can in principle be synthesized by value or policy iteration. However, if the system dynamics and the reward function are unknown, a learning agent must discover an optimal controller via direct interaction with the environment. Such interactive data gathering commonly leads to divergence towards dangerous or uninformative regions of the state space unless additional regularization measures are taken. Prior works proposed bounding the information loss measured by the Kullback–Leibler (KL) divergence at every policy improvement step to eliminate instability in the learning dynamics. In this paper, we consider a broader family of f-divergences, and more concretely α-divergences, which inherit the beneficial property of providing the policy improvement step in closed form at the same time yielding a corresponding dual objective for policy evaluation. Such entropic proximal policy optimization view gives a unified perspective on compatible actor-critic architectures. In particular, common least-squares value function estimation coupled with advantage-weighted maximum likelihood policy improvement is shown to correspond to the Pearson χ2-divergence penalty. Other actor-critic pairs arise for various choices of the penalty-generating function f. On a concrete instantiation of our framework with the α-divergence, we carry out asymptotic analysis of the solutions for different values of α and demonstrate the effects of the divergence function choice on common standard reinforcement learning problems.
Collapse
|
96
|
Pajarinen J, Thai HL, Akrour R, Peters J, Neumann G. Compatible natural gradient policy search. Mach Learn 2019. [DOI: 10.1007/s10994-019-05807-0] [Citation(s) in RCA: 8] [Impact Index Per Article: 1.6] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/29/2022]
|
97
|
Schürmann T, Mohler BJ, Peters J, Beckerle P. How Cognitive Models of Human Body Experience Might Push Robotics. Front Neurorobot 2019; 13:14. [PMID: 31031614 PMCID: PMC6470381 DOI: 10.3389/fnbot.2019.00014] [Citation(s) in RCA: 13] [Impact Index Per Article: 2.6] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/30/2018] [Accepted: 03/21/2019] [Indexed: 01/08/2023] Open
Abstract
In the last decades, cognitive models of multisensory integration in human beings have been developed and applied to model human body experience. Recent research indicates that Bayesian and connectionist models might push developments in various branches of robotics: assistive robotic devices might adapt to their human users aiming at increased device embodiment, e.g., in prosthetics, and humanoid robots could be endowed with human-like capabilities regarding their surrounding space, e.g., by keeping safe or socially appropriate distances to other agents. In this perspective paper, we review cognitive models that aim to approximate the process of human sensorimotor behavior generation, discuss their challenges and potentials in robotics, and give an overview of existing approaches. While model accuracy is still subject to improvement, human-inspired cognitive models support the understanding of how the modulating factors of human body experience are blended. Implementing the resulting insights in adaptive and learning control algorithms could help to taylor assistive devices to their user's individual body experience. Humanoid robots who develop their own body schema could consider this body knowledge in control and learn to optimize their physical interaction with humans and their environment. Cognitive body experience models should be improved in accuracy and online capabilities to achieve these ambitious goals, which would foster human-centered directions in various fields of robotics.
Collapse
|
98
|
Koc O, Peters J. Learning to Serve: An Experimental Study for a New Learning From Demonstrations Framework. IEEE Robot Autom Lett 2019. [DOI: 10.1109/lra.2019.2896466] [Citation(s) in RCA: 7] [Impact Index Per Article: 1.4] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/09/2022]
|
99
|
Brandherm F, Peters J, Neumann G, Akrour R. Learning Replanning Policies With Direct Policy Search. IEEE Robot Autom Lett 2019. [DOI: 10.1109/lra.2019.2901656] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.2] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/10/2022]
|
100
|
Abstract
The last century has seen a large development in diffraction techniques. The time-of-flight neutron diffraction method is now so advanced that it provides high precision results for position and thermal parameters, which are complementary to other diffraction results from X-ray sources. Here we review the history of neutron sources, the difficulties encountered with the time-of-flight technique and an outlook for applications. In this context, we will show the limitations of existing neutron sources and the expected advantages of new spallation neutron sources. An overview of all corrections to be taken into account with the wavelength-sorting technique will be presented as well as actual results, how to overcome such problems, and the special difficulty of integration of three-dimensional Bragg peaks.
Collapse
|