Hybrid learning mechanisms under a neural control network for various walking speed generation of a quadruped robot.

Legged robots that can instantly change motor patterns at different walking speeds are useful and can accomplish various tasks efficiently. However, state-of-the-art control methods either are difficult to develop or require long training times. In this study, we present a comprehensible neural control framework to integrate probability-based black-box optimization (PIBB) and supervised learning for robot motor pattern generation at various walking speeds. The control framework structure is based on a combination of a central pattern generator (CPG), a radial basis function (RBF) -based premotor network and a hypernetwork, resulting in a so-called neural CPG-RBF-hyper control network. First, the CPG-driven RBF network, acting as a complex motor pattern generator, was trained to learn policies (multiple motor patterns) for different speeds using PIBB. We also introduce an incremental learning strategy to avoid local optima. Second, the hypernetwork, which acts as a task/behavior to control parameter mapping, was trained using supervised learning. It creates a mapping between the internal CPG frequency (reflecting the walking speed) and motor behavior. This map represents the prior knowledge of the robot, which contains the optimal motor joint patterns at various CPG frequencies. Finally, when a user-defined robot walking frequency or speed is provided, the hypernetwork generates the corresponding policy for the CPG-RBF network. The result is a versatile locomotion controller which enables a quadruped robot to perform stable and robust walking at different speeds without sensory feedback. The policy of the controller was trained in the simulation (less than 1 h) and capable of transferring to a real robot. The generalization ability of the controller was demonstrated by testing the CPG frequencies that were not encountered during training.

Locomotion Control With Frequency and Motor Pattern Adaptations.

Existing adaptive locomotion control mechanisms for legged robots are usually aimed at one specific type of adaptation and rarely combined with others. Adaptive mechanisms thus stay at a conceptual level without their coupling effect with other mechanisms being investigated. However, we hypothesize that the combination of adaptation mechanisms can be exploited for enhanced and more efficient locomotion control as in biological systems. Therefore, in this work, we present a central pattern generator (CPG) based locomotion controller integrating both a frequency and motor pattern adaptation mechanisms. We use the state-of-the-art Dual Integral Learner for frequency adaptation, which can automatically and quickly adapt the CPG frequency, enabling the entire motor pattern or output signal of the CPG to be followed at a proper high frequency with low tracking error. Consequently, the legged robot can move with high energy efficiency and perform the generated locomotion with high precision. The versatile state-of-the-art CPG-RBF network is used as a motor pattern adaptation mechanism. Using this network, the motor patterns or joint trajectories can be adapted to fit the robot's morphology and perform sensorimotor integration enabling online motor pattern adaptation based on sensory feedback. The results show that the two adaptation mechanisms can be combined for adaptive locomotion control of a hexapod robot in a complex environment. Using the CPG-RBF network for motor pattern adaptation, the hexapod learned basic straight forward walking, steering, and step climbing. In general, the frequency and motor pattern mechanisms complement each other well and their combination can be seen as an essential step toward further studies on adaptive locomotion control.

Generic Neural Locomotion Control Framework for Legged Robots.

In this article, we present a generic locomotion control framework for legged robots and a strategy for control policy optimization. The framework is based on neural control and black-box optimization. The neural control combines a central pattern generator (CPG) and a radial basis function (RBF) network to create a CPG-RBF network. The control network acts as a neural basis to produce arbitrary rhythmic trajectories for the joints of robots. The main features of the CPG-RBF network are: 1) it is generic since it can be applied to legged robots with different morphologies; 2) it has few control parameters, resulting in fast learning; 3) it is scalable, both in terms of policy/trajectory complexity and the number of legs that can be controlled using similar trajectories; 4) it does not rely heavily on sensory feedback to generate locomotion and is thus less prone to sensory faults; and 5) once trained, it is simple, minimal, and intuitive to use and analyze. These features will lead to an easy-to-use framework with fast convergence and the ability to encode complex locomotion control policies. In this work, we show that the framework can successfully be applied to three different simulated legged robots with varying morphologies and, even broken joints, to learn locomotion control policies. We also show that after learning, the control policies can also be successfully transferred to a real-world robot without any modifications. We, furthermore, show the scalability of the framework by implementing it as a central controller for all legs of a robot and as a decentralized controller for individual legs and leg pairs. By investigating the correlation between robot morphology and encoding type, we are able to present a strategy for control policy optimization. Finally, we show how sensory feedback can be integrated into the CPG-RBF network to enable online adaptation.

Error-Based Learning Mechanism for Fast Online Adaptation in Robot Motor Control.

Existing state-of-the-art frequency adaptation mechanisms of central pattern generators (CPGs) for robot locomotion control typically rely on correlation-based learning. They do not account for the tracking error that may occur between the actual system motion and CPG output, leading to the loss of precision, unwanted movement, inefficient energy locomotion, and in the worst cases, motor collapse. To overcome this problem, we developed online error-based learning for frequency adaptation of CPGs. The learning mechanism used for error reduction is a novel modification of the dual learner (DL) called dual integral learner (DIL). Being able to reduce tracking and steady-state errors, it can also perform fast and stable learning, adapting the CPG frequency to match the performance of robotic systems. Control parameters of the DIL are more straightforward for complex systems (like walking robots), compared to traditional correlation-based learning, since they correspond to error reduction. Due to its embedded memory, the DIL can relearn quickly and recover spontaneously from the previously learned parameters. All these features are not covered by the existing frequency adaptation mechanisms. We integrated the DIL into a neural CPG-based motor control system for use on different legged robots with various morphologies for evaluation. The results show that: 1) the DIL does not require precise adjustment of its parameters to fit specific robots; and 2) the DIL can automatically and quickly adapt the CPG frequency to the robots such that the entire trajectory of the CPG can be precisely followed with very low tracking and steady-state errors. Consequently, the robots can perform the desired movements with more energy-efficient locomotion compared to the state-of-the-art correlation-based learning mechanism called frequency adaptation through fast dynamical coupling (AFDC). In the future, the proposed error-based learning mechanism for fast online adaptation in robot motor control can be used as a basis for trajectory optimization, universal controllers, and other studies concerning the change of intrinsic or extrinsic parameters.