In this paper, we discuss the design of a parallel-snake model for lane detection and the use of a Kalman filter for tracking. The parallel-snake model is an extension of the open active contour model through the application of a parallel constraint to two open snakes. Compared with other models, this model can handle lanes with broken boundaries and reduce the convergence time with the aid of the parallel constraint and a double external energy force from two parallel snakes. To solve the problem in previous snake models, whereby the external force is lost on images with a low gradient, a balloon force is utilized to expand the double snakes from the center of the road to the lane boundaries. Because lane boundaries do not retain the parallel property, the captured images are transformed into a bird's-eye view to retrieve the parallel property of lane boundaries by planar homography. At least four corresponding points are determined and the EM-based vanishing point estimation algorithm is applied to these points to estimate the planar homography. Finally, we use a Kalman filter for parameter optimization in lane tracking considering the continuity of lane parameters between consecutive frames; i.e., to predict the parameters of subsequent frames from the previous frame and refine the estimated results to improve robustness. Experimental results show that the proposed method achieves good performance on lane datasets with shadows, variations in illumination, and broken boundaries. Furthermore, it can handle both structured and unstructured (country) roads well.
The design of feedback controllers for bipedal robots is challenging due to the hybrid nature of its dynamics and the complexity imposed by high-dimensional bipedal models. In this paper, we present a novel approach for the design of feedback controllers using Reinforcement Learning (RL) and Hybrid Zero Dynamics (HZD). Existing RL approaches for bipedal walking are inefficient as they do not consider the underlying physics, often requires substantial training, and the resulting controller may not be applicable to real robots. HZD is a powerful tool for bipedal control with local stability guarantees of the walking limit cycles. In this paper, we propose a non traditional RL structure that embeds the HZD framework into the policy learning. More specifically, we propose to use RL to find a control policy that maps from the robot's reduced order states to a set of parameters that define the desired trajectories for the robot's joints through the virtual constraints. Then, these trajectories are tracked using an adaptive PD controller. The method results in a stable and robust control policy that is able to track variable speed within a continuous interval. Robustness of the policy is evaluated by applying external forces to the torso of the robot. The proposed RL framework is implemented and demonstrated in OpenAI Gym with the MuJoCo physics engine based on the well-known RABBIT robot model.
We investigated multi-finger synergies stabilizing the total moment of force and the total force when the subjects produced a quick cyclic change in the total moment of force. The seated subjects performed the task with the fingers of the dominant arm while paced by the metronome at 1.33 Hz. They were required to produce a rhythmic, sine-like change in the total pronation-supination moment of force computed with respect to the midpoint between the middle and ring fingers. The framework of the uncontrolled manifold hypothesis was used to compute indices of stabilization of the total moment and of the total force across 20 cycles. Variance of the total moment showed a cyclic pattern with peaks close to the peak rate of the moment change. Variance of the total force was maximal close to peak moment into supination. Higher magnitudes of the moment directed against the required moment direction (antagonist moment) were produced by individual fingers during supination efforts as compared to pronation efforts. Indices of multi-finger synergies showed across-trials stabilization of the total moment over the whole cycle but not of the total force. These indices were smaller during supination efforts. We conclude that the central nervous system facilitates multi-finger synergies stabilizing the total rotational action across a variety of tasks. Synergies stabilizing the total force are not seen in tasks that do not explicitly require accurate force control. Pronation efforts are performed more efficiently and with better stabilization of the action.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.