Multi-layer feed-forward neural network for estimating a 6DOF robot's internal joint angles from cartesian coordinates. The neural network is modeled around the concept of how transformation and coordinate matrices feedforward to estimate the angle of each internal joint.
Requirements:
python
version 3.7.6 or laterconda
version 3.18.11 or latertorch
version 1.7.1 or later,torchvision
, and 'pytorch-pfn-extras' To install:git clone github.com/alexander-jh/KineNet.git
conda create --name <env_name> --file requirements.txt
Execute command from conda environment.
python3 src/train.py
python3 src/SimDataGenerator.py <# of batches to generate>
.
Specify type of data to use:
python 3 src/test.py
Over the training cycle there's an approximate loss of around 10% on average across all 6 joints.
Looking at results by joint, the T-joint is the only joint that's consistently inaccurate.
Mean difference in radians between joint estimation and actual.
S-Joint | L-Joint | U-Joint | R-Joint | B-Joint |
---|---|---|---|---|
-0.0035 | 0.0075 | 0.0042 | -0.0129 | 0.0036 |