Jeffrey N. Shelton

Efficiently Generating the Ballistic Phase of Human-Like Aimed Movement

Human-like movement can be efficiently generated by driving a damped inertial plant toward a fixed target with position-based control, then switching to conventional feedback control as the desired endpoint is approached. Key characteristics of human motion, not easily replicated with traditional control architectures, result from tracking a position-based actuation template derived from human trials. Computation of the applied forcing function requires only linear scaling of this template, or displacement-normalized actuation program (DNAP). Simulated ballistic movements generated with the proposed method are shown to be consistent with human subject kinematic trajectories.

Single Feedback Model of Human Goal-Directed Movement

Two prominent models frequently used to explain targeted human movement are the stochastic optimized-submovement model and the minimum variance model. Both successfully explain the speed-accuracy tradeoff known as Fitts’ law, but neither is complete. The former cannot predict movement trajectory between the endpoints, while the latter is not congruent with the multiple movement segments often observed in human motion. In this paper, a new model is proposed in which an aimed movement consists of two submovements and a single feedback instant, with the trajectory of each submovement being individually optimized. Simulations using the proposed model show that the optimal transition between two submovements occurs at an early stage of the movement, and produces a sharp peak in the acceleration profile. This result is consistent with psychophysical data. Also observed in numerical simulation is the bell-shaped positional variance curve that is in agreement with psychophysical data.Copyright

Robust, Non-Overshooting Control of a Rigid Rotating Link Using Exponential Segmentation

Exponential pyramid architecture plays a key role in human visual processing, as well as in learned muscle movement. Attempting to mimic the human capacity for making rapid minimally-overshooting movements, a variant of exponential pyramid architecture, called exponential segmentation, is utilized to develop a methodology for controlling the response of rigid links to step inputs. The resulting method consists of issuing iterative open-loop commands, each of which are guaranteed not to overshoot the desired target. Each time the mechanism comes to a stop, a new open loop command is issued that again drives the mechanism toward the desired final position without overshoot. Since this method always undershoots the target, the process is repeated until the link is within a specified interval of the desired final position. Using this framework, a designer can easily make tradeoffs between response and robustness, as this method is robust with respect to input perturbations, pivot friction, link inertia and damping rates. Furthermore, this method can be implemented using positional measurement alone.Copyright

Smoothly Transitioning Between Ballistic and Corrective Control to Produce Human-Like Movement

Human reaching movement appears to consist of an initial ballistic segment that drives the hand toward the target, then a corrective segment that brings the hand into the target region. This article discusses how the motions produced by two different controllers, one guiding the ballistic portion and one directing the corrective potion, can be merged into a single smooth movement that is reminiscent of human reaching. Simulated movements based on the proposed methodology are shown to be consistent with human kinematic trajectories.Copyright

Efficient generation of human-like kinematics in the ballistic phase of point-to-point movement

Human movement is often considered to have two phases: a ballistic phase that brings the limb near the target, and a corrective phase that locates the limb on the target. This article proposes the use of a single non-dimensional curve that can be scaled to produce human-like ballistic movement across a wide range of reaching task configurations. Control of the scaled forcing function requires only that a look-up table of input levels be referenced in response to position measurements. Importantly, the proposed method accurately accounts for the significant kinematic variations evident between individual trials. In comparison, typical feedforward methods replicate the desired mean behavior, but fail to generate the requisite bell-shaped rise and fall in positional variance. Stochastic-optimal feedback control produces the appropriate velocity and variance curves, but carries with it a high computational burden.

Rapid Computation of Time-Optimal, Open-Loop Forearm Movement

Abstract Minimal movement time for open-loop rotation of the human forearm, along with the associated input signal to the forearm muscles, is calculated using matrix multiplication. This permits rapid evaluation of movement times across a four-dimensional mesh of initial conditions, each moving to a common terminal state. The described discrete-time solution is based on the continuous-time solution of Tanaka et al., and the minimum-variance theory of Harris and Wolpert. Underlying algorithm concepts are discussed, and proofs of solution existence are provided.