Developing and validating reactive control for intelligent robot behaviors on the Robotrek platform

Аннотация

The present paper fills a knowledge gap that crucially exists on the imperfectly empirical knowledge of how the three basic intelligent behaviours (environmental adaptation, obstacle avoidance and path planning) scale, in terms of computational demands and degrees and speeds of success as applied to any given robot (using the publicly available Robotrek educational robotic kit (based on Tracduino microcontrollers, Robotrack IDE)). A reactive control software extension was implemented and tested with simulation and in experimentally physical tests. Among the core conclusions, it is possible to notice that the platform showed a high level of efficacy in structured settings: 92% success rate of avoiding obstacles at a threshold of 15 cm with a latency of less than 200 ms, and straight-line tracking within <2 cm average deviation. These findings confirm the potential of Robotrek in carrying out fundamental autonomous behaviours and serves to offer empirical benchmarks to reactive control paradigms in a resource-constrained hardware. Despite this, the study shows inherent shortcomings the research team including: limited ability to perceive the environment; failure to handle dynamic adversaries, or to optimize routes; and lack of robustness in unstructured contexts, which represent a sizeable breach between reactive capability and actual autonomous thought. The implications are future work in computer-vision integration (e.g. Raspberry Pi), and embedded AI (reinforcement learning, path planners), as well as energy-aware operation, to progress towards adaptable, deployable, robots.