humanoid-iitk / automi_v1 Goto Github PK

Immediate: Some gazebo humanoid bot to test modules.
Final: Automi's gazebo model.

Blockers:

• AUTOMI model is currently unstable due to friction (hypothesis).

• Find the shortest possible route to the goal.
• Minimize effort.
• Avoid collision with obstacles.

Approaches:

• Greedy approach: Find the next favourable milestone. Optimise an objective.
• Global planner: graph-based planners.

Move hardcoded parameters to config.yml files.

Given: RGBD image
Output: Surface pixels, Equation of ground plane OR point cloud of surface points.

Approaches:

• Generate point cloud -> use RANSAC to get equation of the ground plane,
• Cluster points belonging to same surface.
• Occupancy grid map.

Immediate: Test and integrate vanilla SLAM modules already present on github.
Final: Design a version which is robust to noise.

Add utility functions and convertor classes for msg to objects (and reverse) without replacing the previously present object completely.