Projects
Selected robotics and autonomy research projects.
EARTH
Excavation Autonomy with Resilient Traversability and Handling
EARTH is a framework for autonomous excavators and earth-movers that integrates perception, planning, safe control, hydraulic actuation, and field deployment for embodied construction autonomy.
Safe Control with CLF-CBF-QP
Reactive safe planning for excavator-arm operation
Safe planning mitigates risks associated with human injury, equipment damage, and environmental harm while enforcing constraints on key operating parameters. Control Lyapunov and Control Barrier functions support reactive planners that respond to obstacles before potential collisions occur.
A Trajectory-Level Constraint Aware Framework For Safe Sampling
In progress - trajectory-level constraint-aware safe sampling
This work integrates a single smooth trajectory-level CBF correction into diffusion-based trajectory generation for fast, safety-aware planning in cluttered environments. The abstract examples show safe trajectory generation in Maze2D narrow-passage and PointMass2D dense-obstacle settings.
Status: In progress
Dynamics-Guided Trajectory Generation for Hydraulic Excavators
In progress - diffusion-based excavator trajectories with learned dynamics guidance
This ongoing work studies diffusion-based trajectory generation for hydraulic excavators with learned forward and inverse dynamics models. The diffusion model is to be trained on demonstrations simulated using the forward and inverse dynamics models. The sampler uses gradients from learned forward dynamics models for guidance during diffusion sampling to produce observation-control trajectories that are more dynamically coherent.
Status: In progress