Abstract
Introduction
Autonomous robots will revolutionize the world we live in. Few industries will not be completely transformed with the development and adoption of autonomous robotics. However, widespread deployment requires progress both in terms of technical capabilities and societal acceptance and adoption of these robots. My capstone project aims to address both these challenges. The technical component involved developing a modular autonomy stack for Clearpath Jackal robots with the goal of helping bridge the gap between simulation-based research and real-world deployment. The STS component on the other hand investigates how various factors impact societal adoption of autonomous robots. Taken together, the two components of my capstone project are a step in advancing autonomous robotics.
Technical Discussion
The technical portion of my thesis produced a complete, modular autonomy stack implemented in ROS 2 for the Clearpath Jackal robotic platform. The goal of the codebase is to address a gap in robotics research between algorithms tested in simulation, and those deployed on real robots. This gap exists for several reasons, including but not limited to, sensor noise, computational constraints, and non-deterministic conditions in the real-world, but also because of the need for a full autonomy stack to test just a single module on a real robot. A modular autonomy stack helps researchers plug and play with their components, allowing for fast and seamless testing on real robots. My implementation includes four key components. A perception node that processes LiDAR data in real time, filtering floor and ceiling points, and building a 2D occupancy grid of the robot’s surroundings. A planner node that receives this occupancy grid along with the robot’s current and goal position, adds a 0.5-meter safety buffer around obstacles to account for the robot’s footprint, converts the environment into a graph, and applies the A* algorithm to find collision-free paths. A controller node that implements the pure-pursuit control algorithm, continuously computing velocity commands to track the planned trajectory. Finally, an autonomy node that supervises the robot’s state, implementing safety features including switching between autonomous and teleoperated control modes, and an emergency stop that immediately stops the robot, overriding all other commands. The architecture was designed with modularity in mind, for the explicit purpose of simplifying future research. This work establishes a foundation for advancing robotics research on real hardware.
STS Discussion
My STS research investigates what factors influence whether people and organizations choose to adopt autonomous robotic systems. Technical capability alone does not guarantee acceptance. The most obvious example of this is that users must have confidence that a system is safe in order to adopt it. My research analyzes how adoption of these systems is shaped by four categories of factors: human factors like prior experience and familiarity, robot factors including performance reliability and design transparency, environmental factors such as organizational norms and task context, and institutional factors like regulatory clarity and assignment of responsibility. There are real-world cases that demonstrate these factors impacting adoption. For example, autonomous vehicles have been met with a lot of skepticism in the public despite technological progress, in large part due to lack of design transparency and regulatory frameworks that are still underdeveloped. On the other hand, warehouse robots have achieved broad adoption in controlled environments with well-defined task boundaries and human-robot roles. Healthcare robots lie somewhere in the middle. Despite being capable, they have faced adoption barriers because of questions over who is responsible and a lack of clear regulation. My findings show that successful autonomous robot deployment requires not just capable technology, but also depends on achieving various key factors ranging from user familiarity and robot reliability to design transparency and institutional mechanisms.
Conclusion
There are two main challenges to widespread adoption of autonomous robots. From a technical standpoint, algorithmic advances are required, particularly in the algorithms deployed in the real-world as opposed to in simulation. These systems then also need to be accepted societally. This thesis addressed these two challenges. The Jackal autonomy stack helps reduce the barriers to real-world research by allowing researchers to experiment on real robotic systems with minimal setup required. The STS research helps create a better understanding of what is required for widespread adoption of these robots. Together, the two components of this project advance the deployment of autonomous robotic systems that are both technically capable and socially accepted.
