Individual Goal Setting

Role Focus: Algorithm & AI Engineer

Expectations on the Project

  1. Technical contribution: I expect to lead the design, implementation and optimisation of the "brain" part of the project, i.e. the stabilisation control algorithm. I hope to develop a set of efficient and robust algorithms that can not only stabilise the platform, but also intelligently adapt to different water environments and user behaviours.
  2. Results Expectation: I expect the final prototype to significantly enhance the user experience, and its stability effect can be clearly reflected from the test data and on-water demonstration, and become the most core technical highlight of the project.
  3. Team Collaboration: I expect to work closely with the hardware engineers to ensure the accuracy of sensor data acquisition and real-time actuator control; and with the testers to transform the field test data into the basis for algorithm optimisation.

Learning Gaps Identified

  1. Theory-to-practice gap: Although control theory (e.g., PID) has been learnt, there is a lack of practical experience in applying it to multiple-input multiple-output (MIMO) nonlinear systems (multiple thrusters acting together). In-depth study of modern control theory (e.g., LQR) in water robotics is needed.
  2. Sensor fusion techniques: Knowledge of the theory of Kalman Filter and its variants (e.g., complementary filtering), but lack of experience implementing and debugging in code (C++/Python) on embedded platforms to obtain accurate attitude angles.
  3. Combination of AI and Control: Interested in machine learning, but unsure how to combine AI techniques such as Reinforcement Learning or Neural Networks with traditional control methods for predictive control or adaptive parameter tuning. This is a cutting edge area that I hope to explore.
  4. Programming for Real-Time Systems: Lack of experience writing critical control loops in Real-Time Operating Systems (RTOS) or high priority threads to ensure time-sensitive control.

My Responsibilities in the Project

  1. Algorithm Research and Selection: Research and determine the most suitable control architecture (e.g. centralised PID, decentralised control, model-based control).

  2. Sensor Data Fusion: Responsible for writing code to fuse IMU (gyroscope, accelerometer) and other data to calculate the current accurate Roll, Pitch and Yaw angles of the device.

  3. Core Control Algorithm Development: Design and implement control algorithms to convert the deviation of attitude angle from the target angle (usually horizontal) into thrust commands for each thruster.

  4. SIMULATION AND TESTING: Perform algorithm simulation using MATLAB/Simulink or Python prior to physical prototype. Once the prototype is out, responsible for writing test scripts, collecting data, analysing algorithm performance and continuous iterative optimisation.

  5. AI Functionality Exploration: After the basic functionality has been implemented, explore advanced functionality based on machine learning, such as:

    ① Learning the habitual weight distribution of a specific user;

    ② Identifying different types of wave patterns and performing predictive compensation.

ISDN 4002 Individual Scope-Tony