Farbod Fahimi

Online human training of a myoelectric prosthesis controller via actor-critic reinforcement learning

As a contribution toward the goal of adaptable, intelligent artificial limbs, this work introduces a continuous actor-critic reinforcement learning method for optimizing the control of multi-function myoelectric devices. Using a simulated upper-arm robotic prosthesis, we demonstrate how it is possible to derive successful limb controllers from myoelectric data using only a sparse human-delivered training signal, without requiring detailed knowledge about the task domain. This reinforcement-based machine learning framework is well suited for use by both patients and clinical staff, and may be easily adapted to different application domains and the needs of individual amputees. To our knowledge, this is the first my-oelectric control approach that facilitates the online learning of new amputee-specific motions based only on a one-dimensional (scalar) feedback signal provided by the user of the prosthesis.

Real-time obstacle avoidance for multiple mobile robots

An efficient, simple, and practical real time path planning method for multiple mobile robots in dynamic environments is introduced. Harmonic potential functions are utilized along with the panel method known in fluid mechanics. First, a complement to the traditional panel method is introduced to generate a more effective harmonic potential field for obstacle avoidance in dynamically changing environments. Second, a group of mobile robots working in an environment containing stationary and moving obstacles is considered. Each robot is assigned to move from its current position to a goal position. The group is not forced to maintain a formation during the motion. Every robot considers the other robots of the group as moving obstacles and hence the physical dimensions of the robots are also taken into account. The path of each robot is planned based on the changing position of the other robots and the position of stationary and moving obstacles. Finally, the effectiveness of the scheme is shown by modeling an arbitrary number of mobile robots and the theory is validated by several computer simulations and hardware experiments.

Robust control of underactuated bipeds using sliding modes

The purpose of this paper is to present a robust tracking control algorithm for underactuated biped robots capable of self-balancing in the presence of external disturbances. The biped is modeled as a five-link planar robot with four actuators located at hip and knee joints. A sliding mode control law has been developed for the biped to follow a human-like gait trajectory while keeping the torso nearly upright. The control forces are calculated by defining four first-order sliding surfaces as a linear combination of the torso and the four joint tracking errors. The control approach is shown to guarantee that all trajectories will reach and stay on these surfaces during each step, while the walking cycle stability is maintained through a Lyapunov function. The criteria for asymptotic stability of the surfaces are presented and a numerical search method is implemented for the selection of the corresponding surface parameters. The paper further investigates the robustness of the controller in response to disturbances. Numerical simulations demonstrate the tracking stability of the bipeds multistep walk and its human-like response to an external disturbance.

An improved inverse kinematic and velocity solution for spatial hyper-redundant robots

A new and efficient kinematic position and velocity solution scheme for spatial hyper-redundant manipulators is presented. The manipulators arm has discrete links and universal joints. Backbone curve concepts and a modal approach are used to resolve the manipulators redundancy. The effects of the mode shapes and the slope of the backbone curve at the starting point on the workspace are studied. It is shown that the usage of conventional mode shapes limits the workspace of the hyper-redundant arm. By introducing new mode shapes, an improved workspace is obtained. A simple and efficient recursive fitting method is introduced to avoid complications involved with solving systems of nonlinear algebraic equations. This method also guarantees the existence of solutions for the inverse kinematic problem at the velocity level. Velocity properties of the backbone curve are investigated and the inverse velocity propagation is solved for the spatial hyper-redundant arm. The velocity propagation scheme is recursive and is efficiently applicable to any number of links.

The Development of a Myoelectric Training Tool for Above-Elbow Amputees

The objective of above-elbow myoelectric prostheses is to reestablish the functionality of missing limbs and increase the quality of life of amputees. By using electromyography (EMG) electrodes attached to the surface of the skin, amputees are able to control motors in myoelectric prostheses by voluntarily contracting the muscles of their residual limb. This work describes the development of an inexpensive myoelectric training tool (MTT) designed to help upper limb amputees learn how to use myoelectric technology in advance of receiving their actual myoelectric prosthesis. The training tool consists of a physical and simulated robotic arm, signal acquisition hardware, controller software, and a graphical user interface. The MTT improves over earlier training systems by allowing a targeted muscle reinnervation (TMR) patient to control up to two degrees of freedom simultaneously. The training tool has also been designed to function as a research prototype for novel myoelectric controllers. A preliminary experiment was performed in order to evaluate the effectiveness of the MTT as a learning tool and to identify any issues with the system. Five able-bodied participants performed a motor-learning task using the EMG controlled robotic arm with the goal of moving five balls from one box to another as quickly as possible. The results indicate that the subjects improved their skill in myoelectric control over the course of the trials. A usability survey was administered to the subjects after their trials. Results from the survey showed that the shoulder degree of freedom was the most difficult to control.

Myoelectric training systems

Myoelectric prostheses aim to restore the functionality of amputated limbs and improve the quality of life of amputees. Myoelectric training systems are used to train and assess the ability of amputees in how to use myoelectric technology in advance of receiving their actual myoelectric prostheses. This article describes the myoelectric training systems that have been developed over the last 20 years in both the literature and commercial industries. The results of this article will identify common features in the training systems and suggest areas for future improvement.

Full formation control for autonomous helicopter groups

This paper reports the design of sliding-mode control laws for controlling multiple small-sized autonomous helicopters in arbitrary formations. Two control schemes, which are required for defining arbitrary three-dimensional formation meshes, are discussed. In the presented leader–follower formation control schemes, each helicopter only needs to receive motion information from at most two neighboring helicopters. A nonlinear six-degree-of-freedom dynamic model has been used for each helicopter. Four control inputs, the main and the tail rotor thrusts, and the roll and pitch moments, are assumed. Parameter uncertainty in the dynamic model and wind disturbance are considered in designing the controllers. The effectiveness and robustness of these control laws in the presence of parameter uncertainty in the dynamic model and wind disturbances are demonstrated by computer simulations.

Optimal Adaptive Active Suspensions for a Full Car Model

SUMMARY In order to present a useful method for designing active suspension of a vehicle, a linear full-car model is used in this investigation. In this model, the dampers of passive system are totally replaced by actuators. The actuators are controlled with optimal full state vector feedback. After determining feedback coefficients, the responses of active and passive systems were compared and it was found that performance of active system is much superior. It is desired that, changes in vehicle parameters would not affect the systems performance and hence should not violate its optimality. In other words, the system should behave adaptively using Model Reference Adaptive Control. The optimally controlled active suspension was used as a model for the active suspension of vehicle. In this way, the suspension of vehicle is controlled in such a way that its output approaches to that of the optimal active model. Thus the suspension should behave just like the optimal one.

Full drive-by-wire dynamic control for four-wheel-steer all-wheel-drive vehicles

Most of the controllers introduced for four-wheel-steer (4WS) vehicles are derived with the assumption that the longitudinal speed of the vehicle is constant. However, in real applications, the longitudinal speed varies, and the longitudinal, lateral, and yaw dynamics are coupled. In this paper, the longitudinal dynamics of the vehicle as well as its lateral and yaw motions are controlled simultaneously. This way, the effect of driving/braking forces of the tires on the lateral and yaw motions of the vehicle are automatically included in the control laws. To address the dynamic parameter uncertainty of the vehicle, a chatter-free variable structure controller is introduced. Elimination of chatter is achieved by introducing a dynamically adaptive boundary layer thickness. It is shown via simulations that the proposed control approach performs more robustly than the controllers developed based on dynamic models, in which longitudinal speed is assumed to be constant, and only lateral speed and yaw rate are used as system states. Furthermore, this approach supports all-wheel-drive vehicles. Front-wheel-drive or rear-wheel-drive vehicles are also supported as special cases of an all-wheel-drive vehicle.

Non-Iterative nonlinear model predictive approach applied to the control of helicopters' group formation

Two geometrical formation schemes that allow the definition of any desired three-dimensional formation mesh for a group of helicopters are presented. Each formation scheme, which defines the leader-follower geometry of the formation mesh, has four parameters. These formation parameters are directly used as the output of decentralized controllers that independently control each helicopter in the group. The decentralized controllers are designed using a non-iterative Nonlinear Model Predictive Control (NMPC) method. The Continuation method is used for solving, in real-time, for future control actions that minimize a NMPC cost function. It is shown by analyzing the number of floating point operations per calculation cycle that the calculation load of the NMPC method for this application is quite manageable for todays industrial embedded computers. Simulations show that the formation schemes along with the NMPC controller can initialize and keep the formation of a group of helicopters even in the presence of bounded parameter uncertainty and environmental disturbance.