Neural Control, Memory, and Learning for Complex Behaviors in Multi Sensori-Motor Robotic Systems
Final Report Abstract
Living creatures, like walking animals, have found fascinating solutions for the problem of complex locomotion control. Their movements show the impression of elegance including versatility, energy efficiency, and adaptivity. During the last few decades, roboticists have tried to imitate such natural properties with artificial legged locomotion systems by using different approaches. Several of them have focused on biomechanics including bio-inspired structures, artificial muscles, and advanced materials. Others have concentrated on locomotion control mechanisms based on machine learning algorithms, classical engineering control techniques, and biologically-inspired control mechanisms. However, their levels of performance are still far from the natural ones. By contrast, animal locomotion principles seem to largely depend on not only biomechanics (like structures, muscles, and materials) but also neural control, memory, and learning mechanisms with sensory feedback. Inspired by this, the project focuses on developing biomechanics and different neural mechanisms with sensory feedback for a multi sensori-motor robotic system (i.e., walking robot) to approach the living creatures in their level of performance. Specifically, we have developed: I. Biomechanical hexapod robot with bio-inspired segmented leg and body structures and a special biological material, II. Antagonistic joint control for muscle-like behavior generation and energy-efficient motion, III. Adaptive neural forward models for sensory prediction and state estimation, IV. Neural central mechanisms (central pattern generators, CPGs) with sensory feedback for complex movement generation, V. Neural synaptic plasticity (long-term memory) for learning and adaptation, VI. Neural temporal memory for temporary memorization, VII. Neural predictive control for goal-directed navigation learning. Combining all these components and implementing them on the developed hexapod robot lead to autonomous complex behaviors. The behaviors include adaptive walking on difficult terrains, a multitude of insect-like gaits, energy-efficient walking, adaptive climbing over high obstacles (up to 85% of leg length), crossing a large gap (approx. 44% of body length), malfunction compensation, adaptive obstacle avoidance in very complex environments, memory-guided navigation, and goal-directed behavior. Additionally, the developed components (II-VII) are general and transferable to different robotic and non-robotic systems. There are a diverse set of applications and technologies, to which the results of this project can potentially contribute. Examples include using our antagonistic joint control (muscle model) for generating compliance of rehabilitation, prosthesis, orthosis, or robot arm systems, using our neural central mechanisms with sensory feedback for generating adaptive walking of multi-legged robots, using our neural learning and memory for obtaining adaptivity and robustness of behaving robotic systems. Apart from these examples, our developed robot system shows the possibility of using biological inspiration for developing a new advanced robotic technology, which is capable of operating in complex environments. We believe that the robot system will give the possibility to computational neuroscientists and biologists to study biomechanical and neural functions with a real physical system for resolving new scientific questions. The system can be also used for education as well as real-world applications, like search and rescue, inspection, and planetary exploration. The results of this project were successfully disseminated through public lectures, scientific events, and Hanover Fair (see http://manoonpong.com/events.html).
Publications
- (2013) Combining Correlation- Based and Reward-Based Learning in Neural Control for Policy Improvement. Advances in Complex Systems (Advs. Complex Syst.), Volume 16, Issue 02n03
Manoonpong, P.; Kolodziejski, C.; Wörgötter, F.; Morimoto J.
(See online at https://doi.org/10.1142/S021952591350015X) - (2013) Information Dynamics based Self-Adaptive Reservoir for Delay Temporal Memory Tasks. Evolving Systems, 4(4): 235-249
Dasgupta, S.; Wörgötter, F.; Manoonpong, P.
- (2013) Neural Control and Adaptive Neural Forward Models for Insect-like, Energy-Efficient, and Adaptable Locomotion of Walking Machines. Front. Neural Circuits 7: 12
Manoonpong, P.; Parlitz, U.; Wörgötter F.
(See online at https://doi.org/10.3389/fncir.2013.00012) - (2014) Biologically Inspired Modular Neural Control for a Leg-Wheel Hybrid Robot. Advances in Robotics Research (ARR), Techno Press, Vol. 1, No. 1, 101-126
Manoonpong, P.; Wörgötter, F.; Laksanacharoen, P.
(See online at https://dx.doi.org/10.12989/arr.2014.1.1.101) - (2014) Biologically-Inspired Adaptive Obstacle Negotiation Behavior of Hexapod Robots. Front. Neurorobot. 8:3
Goldschmidt, D.; Wörgötter, F.; Manoonpong, P.
(See online at https://doi.org/10.3389/fnbot.2014.00003) - (2014) Multiple Chaotic Central Pattern Generators with Learning for Legged Locomotion and Malfunction Compensation. Information Sciences
Ren, G.; Chen, W.; Dasgupta, S.; Kolodziejski, C.; Wörgötter, F.; Manoonpong, P.
(See online at https://doi.org/10.1016/j.ins.2014.05.001) - (2014) Neuromechanical Control for Hexapedal Robot Walking on Challenging Surfaces and Surface Classification. Robotics and Autonomous Systems, vol. 62, no. 12, pp. 1777 – 1789
Xiong, X.; Wörgötter, F.; Manoonpong, P.
(See online at https://doi.org/10.1016/j.robot.2014.07.008) - (2014) Virtual Agonist-antagonist Mechanisms Produce Biological Muscle-like Functions: An Application for Robot Joint Control. Industrial Robot: An International Journal, vol. 41, no. 4, pp. 340 – 346
Xiong, X.; Wörgötter, F.; Manoonpong, P.
(See online at https://doi.org/10.1108/IR-11-2013-421) - . (2014) Neuromodulatory Adaptive Combination of Correlation-based Learning in Cerebellum and Reward-based Learning in Basal Ganglia for Goal-directed Behavior Control. Front. Neural Circuits 8:126
Dasgupta, S.; Wörgötter, F.; Manoonpong, P.
(See online at https://doi.org/10.3389/fncir.2014.00126) - Learning and Chaining of Motor Primitives for Goal-directed Locomotion of a Snake-Like Robot with Screw-Drive Units, International Journal of Advanced Robotic Systems, Volume: 12 issue: 12, December 1, 2015
Chatterjee, S.; Nachstedt, T.; Tamosiunaite, M.; Wörgötter, F.; Manoonpong, P.; Enomoto, Y.; Ariizumi, R.; Matsuno, F.
(See online at https://doi.org/10.5772/61621) - Adaptive and Energy Efficient Walking in a Hexapod Robot under Neuromechanical Control and Sensorimotor Learning. IEEE Transactions on Cybernetics, vol. 46, no. 11, pp. 2521-2534, Nov. 2016
Xiong, X.; Wörgötter, F.; Manoonpong, P.
(See online at https://doi.org/10.1109/TCYB.2015.2479237)
