Welcome to Zili's Website!
Research
Projects
Lyapunov Neural Network
We propose a monotonic Lyapunov Neural Network with inherit properties, to synthesize controllers in nonlinear systems with stability guarantees. The conditions are satisfied by formulating Mixed Integer Linear Program in network training. Ongoing work includes concurrent region of attraction search.
Task-Driven Navigation
Given a robot equipped with a long-range 2-D laser scanner and a short-range exit-door detector (implemented, for instance, using a camera), we propose a method that uses the structure of the local environment to predict the location of the exit of the building. We first trained an auto-encoder deep neural network to predict the heat map of the exit location. This prediction is then used for deciding where the robot should move next and what measurements to acquire.We compare our method against the more traditional frontier-based exploration on floor plans of residential buildings, and demonstrate reductions in both total path length and number of measurements.
Bearing-based Formation Control
Given multiple mobile robots with single integrator dynamics, we designed a bearing-based controller that can steer the system of robots from random positions to a specific shape while minimizing the trajectory length. We first simplified sufficient conditions on global convergence of a bearing-based formation controller. Based on that, we formulated an optimization problem to automatically tune the controller to minimize trajectory lengths. The resulting controller generalizes well to new initial conditions and topologies.
Neural-Network-based 2D SLAM
We used ResNet to simultaneously find relative pose (scan matching) and detect loop closure given two keyframe measurements, from multi-task training. After that, we applied Pose Graph Optimization to correct the pose estimation. To train the model, we collected data from ROS using RRT exploration package in multiple environments.
Semi-automatic Micropipette Aspiration for Biological Cells
ADD