A rational agent for a micromouse

Brief on the Technological Architecture Described in the IET NW Midlands webinar

The architecture described in the “Ethics of Autonomous Vehicles ” webinar focuses on autonomous systems that make decisions independently. The description emphasizes trust, transparency, and ethical decision-making.  

Transparency can be achieved by exposing the decision-making process, this allows an audience to understand why a decision was made and what thoughts went into making that choice. Verifiability is ensured by defining formal specifications such as the road laws in the specific country or a project such as the Consultation on a Regulatory Framework of Automated Vehicles, Law Commission, 2021: Automated Vehicles | Law Commission which is mentioned in the talk. The architecture combines high-level decision-making with reactive components, such as low-level control systems. The high-level decision-making handles ethics and rules, whereas the reactive components monitor real-time interactions with the world. Ethical principles are coded into systems using ethical ordering, e.g. a driverless car prioritising human life over property.

How This Architecture Applies to a Driverless Car:

Driverless cars are complex autonomous systems that rely on sensors for perception. Including cameras, LiDAR and radars to detect other cars, pedestrians and road signs. They also rely on planning, which decides the car’s path, route planning and lane changes, as well as control which dictates steering, acceleration and braking. Transparency is crucial for building trust with passengers and regulatory institutions, the designers should be able to explain the car’s decisions and why it may have stopped in a specific area or why it chose to hit a stationary car instead of a pedestrian. Verifiability ensures the decision-making adheres to traffic rules, such as red lights and giving way to other cars or pedestrians. This can be achieved through formal verification and real-world testing.

How These Ideas Apply to a Micromouse:

Despite the simplicity of a micromouse in comparison to a driverless car, the same principles can be applied where the system is divided into its different functions. Using IR sensors for obstacle avoidance, using light-dependent resistors in conjunction with LEDs for the white line following and ultrasonic sensors for combat. Transparency can be achieved by implementing a logbook or in our case, a website, to record decisions made by the group. Verifiability is ensured via testing, such as experimenting with different values in the code to monitor the micro mouse’s behaviour or testing obstacle avoidance with objects of different sizes and shapes and combat behaviour to ensure it follows set rules. The high-level decision-making (e.g. switching between its different functions or enforcing ethical rules in combat) with the reactive components (e.g. adjusting motor speed to stay on the line) for real-time tasks. Ethical principles can be coded into the micromouse’s decision-making process, such as avoiding collisions instead of completing a task quicker or aligning with set rules for combat.

Conclusion:

The technological architecture discussed can be applied to both complex systems like driverless cars as well as simpler systems like the micromouse. The key principles, of transparency, verifiability and ethical decision-making guarantee autonomous systems are trustworthy, reliable and align with human values. By using these principles we can create a system that not only performs for its intended purpose but also operates safely and can coexist in a world with humans.