Your Name

Assistive Robot

This project will develop effective robotic assistance for individuals with C4-C7 SCI by investigating the appropriate balance between autonomous control and human teleoperation. The research will address three major types of operations in ADL: (1) motion in cluttered environment, (2) motion and force control, and (3) bi-manual (two-arm) robot operations. By integrating evaluation and feedback from users and therapists with technology enhancement (algorithms, information, and interface), our research will systematically improve collaboration between the human users and assistive robots. In addition, insights from human sensorimotor control will guide this assessment-improvement iteration. Our experience and complementary expertise at Tufts (and RPI) and existing equipment/infrastructure will ensure significant, impactful, and rapid progress. The data that we collect and the lessons learned will be used to develop proposals for significant follow-on funding from other sources, such at NIH, NSF, National Institute on Disability, Independent Living, and Rehabilitation Research (NIDILRR) and private foundations.
Report; Videos

Rock Raiders

The project for Intelligent Ground Vehicle Competition(IGVC). It is a kind of self-driving car competition. My task was to read visual data from the ZED camera and develop the lane detection and obstacle avoidance algorithm. In the competition, the lane was white, painting on the grass. The detection accuracy widely varied based on the clearness of the painting and weather conditions. Therefore, I implemented a spatial CNN approach to detect white lines instead of white pixels. For obstacle avoidance, I programmed the rover to obtain the depth data from the ZED camera and searched the obstacles according to the height. After getting the location of the lanes and barriers, I utilized GPS and ROS navigation stack maps, consisting of an obstacle layer and a lane layer, to generate the path, which would guide the vehicle to the desired positions.
Report

Randomness Test For Bridge Game

This research reviews methods of determining whether several random number generators are suitable for Monte Carlo simulations. The simulations are feature transforms of high-dimensional spaces speciﬁcally in the context of simulating Bridge deals. The authors identify two methods for comparing experimental results from a Monte Carlo simulation to either known theoretical probabilities or to results from other simulations. In the research, we apply these methods to several deals from the game Bridge and compare certain statistics from the bridge deals to known theoretical results.
website

Home / Research

Assistive Robot

Rock Raiders

Randomness Test For Bridge Game

Reinforcement Learning for Dam Operation

Contact