Safe navigation using Control Barrier Function

Inspiration

In forestry operations, construction sites, and post-storm cleanup, large volumes of wood waste, such as fallen branches, logs, and debris, must be safely removed from areas where human workers and equipment are present. Manual removal is labor-intensive and exposes workers to safety risks, particularly in cluttered or unstable environments. Autonomous ground robots can assist with debris removal, but they must safely navigate dynamic workspaces while avoiding collisions with obstacles and nearby personnel. The challenge is to design a control framework that enables a robot to follow a designated worker or target location while maintaining safe distances from obstacles, using only local sensing information and without relying on pre-mapped environments. This project addresses that challenge by developing a safety-critical control strategy for robotic navigation and obstacle avoidance.

What it does

This project develops a robotic system for safe removal of wood waste in cluttered environments, with control-theoretic safety guarantees. The approach models a wheeled mobile robot with linear and angular velocity inputs and relies on local sensing to detect nearby obstacles and estimate their relative distance and orientation. Safety is enforced using Control Barrier Functions (CBFs) that maintain a minimum separation distance between the robot and obstacles. The resulting safety constraints are incorporated into a quadratic programming (QP) controller that adjusts the robot’s steering input while tracking a moving target. Simulation results demonstrate safe obstacle avoidance and target following, with future work focusing on multi-obstacle environments and more complex target trajectories.

How we built it

1) Set up the robot and its working area:

The project starts with a mobile robot moving in a flat working area where obstacles such as logs, debris, or other objects may be present. The robot’s job is to move toward and follow a target, such as a worker or a designated pickup location, while safely navigating around objects in its path.

2) Define what “safe” means:

Before the robot moves, the system sets a safe distance that must always be maintained between the robot and any obstacle. This safety zone acts like an invisible protective buffer around both the robot and nearby objects.

3) Continuously check for danger:

As the robot moves, its sensors constantly observe the surroundings and measure how close obstacles are and where they are located relative to the robot. The controller repeatedly checks whether the robot is getting too close to anything and whether its current motion could cause a collision.

4) Adjust motion to stay safe:

When the robot detects that it may get too close to an obstacle, the controller automatically adjusts its turning motion just enough to avoid danger. The goal is not to stop the robot completely, but to gently guide it onto a safer path.

5) Keep following the target:

At the same time, the robot continues to track its intended target, such as a person walking ahead or a desired destination. So the system balances two goals at once: staying safe and moving in the right direction.

6) Make decisions in real time:

During operation, the robot repeatedly performs the same cycle: sensing the environment, checking safety, deciding how to steer, and moving. This happens continuously and quickly, allowing the robot to react immediately when obstacles appear in its path.

7) Test the method in a simulation

The method is first tested in a simulated scenario in which the robot follows a moving target while avoiding an obstacle. The results show that the robot can safely navigate around the obstacle and then return to its path, demonstrating that the approach works in this simple scenario before moving on to more complex situations.

Challenges we ran into

1) Collecting sensor feedback (e.g., camera images) to detect obstacles. 2) Difficulty in connecting with actual UGV hardware for live demonstration.

Accomplishments that we're proud of

1) Development of a Safety-Critical Navigation Framework:

We developed a control framework that enables a mobile robot to navigate cluttered environments while safely following a target. The system ensures that the robot maintains a safe distance from obstacles at all times, reducing the risk of collisions during operation.

2) Integration of Real-Time Obstacle Avoidance:

The project successfully integrates obstacle detection with a control strategy that continuously adjusts the robot’s motion in real time. This allows the robot to react dynamically to obstacles while still progressing toward its goal.

3) Use of Optimization-Based Control:

We implemented an optimization-based control approach that minimally modifies the robot’s steering commands to satisfy safety constraints while preserving the desired motion toward the target. This provides a balance between safety and task performance.

What we learned

Hardware-in-the-loop implementation for safe navigation of a UGV, highlighting the various challenges encountered with the hardware that were absent in simulations.