Vivek Kumar, Total Chaos Robotics Team
Robotics software quality is very crucial to control robot hardware while interacting with the physical world to accomplish a set of tasks. The software needs to tackle sensor errors, software exceptions, robot hardware failures and uncertainties from the external environment. This paper describes the challenges of achieving a culture of software quality for an Android platform robot based on the experiences of the First Tech Challenge (FTC) robotics team, Total Chaos.
The FTC competition requires a three-month turnaround between the competition challenge release and the first competition, and the entire FTC season is about eight months consisting of multiple rounds of robot competitions and performance improvements. To maximize the quality of work in the FTC season, the Charles Handy Athena culture model was employed to provide a small-team culture, working with flexibility, adaptability, and empowerment. The optimum level of power distributions and cooperation was established to maximize team efficiency. The use of the machine learning and agile methodology, embedded with modular software and hardware, further increased the effectiveness of the design process.
The case study of an FTC Android robot demonstrates how the culture of teamwork and collaboration improved the development processes, how Deep Learning TensorFlow based Neural Network enabled accurate autonomous object detection, and how the use of agile development methods drastically improved the efficiency and quality of the Android robot design and implementation process.
By utilizing these processes and models, the team was able to create a robust robot that has the high scoring ability, can handle physical world uncertainties, and has reliable autonomous controls.
Key takeaways include:
- To share the engineering experience and lessons learned while achieving software quality in an Android robotic software-hardware integrated design
- To describe the use of the Charles Handy Athena culture model to establish a solution-oriented, small-team culture, and create a reliable robot when interacting with the complex system in the physical world
- To discuss Deep Learning Neural Network based on TensorFlow framework to improve decision making for autonomous robots
- To explain agile development and modular design enabling timely changes to different parts of the robot without affecting the overall assembly
- By utilizing these processes and models, the team was able to create a robust robot that has the high scoring ability, can handle physical world uncertainties, and has reliable autonomous controls