Closed-Loop Position Control for Growing Robots Via Online Jacobian Corrections

Learning-Based Control Strategies for Soft Robots: Theory, Achievements, and Future Challenges

“…The first work consists of a distal learning method used for designing an open-loop control for a soft manipulator This work learns the inverse static model of a soft manipulator using Deep Deterministic Policy Gradient (DDPG) for end effector path tracking. The latest few works in the field of kinematic control studies the use of differentiable learned models [42] and real-time adaptive models [43]. Differentiable models can reduce the computational effort for estimating control solutions, but is yet to be extended to a more complex system or for fully dynamic control.…”

“…A clear drawback of learning-based models is its inability to be parameterized to design and control variables making Reference Research Problem Achievements Challanges Year [37] Adaptive kinematic controllers Task space control with external disturbances Extension to learning of redundant configurations 2017 and inclusion of trajectory planning [36] Kinematic redundancy resolution with feedback Task space control including orientation with Extension to learning of redundant configurations 2017 control disturbance rejection and inclusion of trajectory planning [40] Visual servoing with adaptive kinematic controllers Task space control based on visual inputs Extension to learning of redundant configurations 2019 and inclusion of trajectory planning [42] Control-oriented quasi-static modelling Tracking of open loop trajectories in the task space Extension to high-dimensional system 2020 [43] Adaptation to morphological changes Tip position and orientation control of a growing robot Inclusion of offline models for control 2021 [39] Multi-objective force and position control Task space control with stiffness optimization Extension to continuous action space 2017 [45] Task-space dynamic control Tracking high dynamics trajectory in the task space Inclusion of feedback for closed-loop control 2017 [46] Task-space model predictive control Model predictive control with a one degree of freedom Extension to high-dimensional system 2018 soft robot [47] Task-space dynamic control with feedback Closed-loop dynamic control for point reaching tasks Extension to tracking tasks 2019 [48] Control-oriented modelling for MPC MPC controller for low dynamics trajectory tracking Extension to high dynamics tracking tasks 2020 [49] Distributed RNNs for data efficiency Real-time dynamic predictive model of a soft manipulator Extension to high-dimensional systems 2021 [50] Hybrid models for data efficiency Model-based optimal control of a soft continuum joint Extension to dynamical high-dimensional systems 2021 [51] Ensemble reinforcement learning for data efficiency Dynamical controller for point reaching tasks in soft arm Extension to trajectory tracking tasks 2021 [44] Dynamic control for locomotion Policy generation of gaits with variation in environmental Extension to faithful simulation environment 2017 conditions [52] Feedback Control with embedded sensors Robust grasping and identification of objects Extension to high-dimensional feedback information 2019…”

Kinematic modeling and motion control of a parallel robotic antenna pedestal

“…Since the actual control requires speed control of the drive motor, it is necessary to solve the velocity relationship equation of the antenna model. The equation is a Jacobian matrix, also known as the first-order motion influence coefficient, which is a mapping between the speed of the operating mechanism and the end speed [17].…”

Model-Based Eversion Control of Soft Growing Robots With Pneumatic Actuation

“…The external force F 3 acts on the system dynamics (2) in a similar way to the damping b. Thus F 3 can either be accounted for by Ω F F + β F in ς as done in (12), or alternatively alongside b in S 23 and in (15), that is…”