Algorithms for AI Robotics - Projects
Summer 2013/2014
Now updated for 2014!
You can solve the projects in groups of 2 people or individually.
Select one of the projects. Then send e-mail to pavel.petrovic@gmail.com to reserve it and get it confirmed, you can propose the implementation details/hw etc. in your mail. After the project is confirmed and specified in more details, register for the lab, if you need. Finally, make a report on project wiki page.
The projects are coming, you can already send an e-mail and register for the time in the lab...
- Omnibot with accellerometer and gyroscope records trajectory and visualizes it on screen.
- LCS for SBOT/Acrob - implement simply Learning Classifier System for one of the simple robot platforms.
- line following robot using camera
- Android platform controlling robot - camera or accellerometer, recording trajectory, image recognition
- learning to detect movements using gyroscopic sensor and ANN
- simulation experiment in Microsoft Robotics Studio
- double pole balancing - a neural network learns to balance two poles on a cart that moves in one dimension (simulation)
- evolution of shapes with Open Dynamics Engine: a shape is encoded in the genotype, the goal is to move it from one place to another with one (or two) motors that oscillate with a specific pattern
- MOEA for robotic task (in simulation)
- all kinds projects in Remotely-Operated laboratory: - automatic docking - multiple robot problem solving with symbolic representation of state - ANN - EA - probabilistic localization/mapping
- POMDP - application for finding solution and visualizing state space and expected reward
- implementation of BBR controller for robot
- robot localization using features in image (landmarks)
- robot localization in arena using distance sensors (US/IR)
- robot localization in changing environment using distance sensors
- using ASUS Xtion for obstacle avoidance
- outdoor mobile robot navigates while avoiding obstacles
- any learning algorithm for robot in real world or simulation
- reinforcement learning - finding path to a goal in maze
- explore the possibilities of ROS and implement simple learning robot