notes

Personal notes
git clone git://git.laack.co/notes.git
Log | Files | Refs

ProbabilisticRobotics.md (2070B)


      1 # Probabilistic Robotics
      2 
      3 ## Links
      4 
      5 ### First Semester
      6 
      7 - [Localization](Localization.md)
      8 - [Histogram Filters](HistogramFilters.md)
      9     - Discrete, Multi-modal, Exponential TC, Approximate
     10 - [Kalman Filters](KalmanFilters.md)
     11     - Continuous, Uni-modal, Quadratic TC (sometimes), Approximate
     12 - [Particle Filters](ParticleFilters.md)
     13     - Continuous, Multi-modal, TC differs, Approximate
     14 - [Bicycle Motion](BicycleMotion.md)
     15 
     16 ### Second Semester
     17 
     18 - Videos 233 - 268 
     19     - Search / Motion Planning
     20         - Shortest Path
     21             - BFS
     22             - [A\*](AStar.md) - uses heuristic function
     23         - Dynamic Programming
     24             - Optimal distance from any location is sometimes useful
     25 - Videos 280 - 312
     26     - Smoothing
     27         - Interpolate between turns to smooth across different steps
     28             - This uses gradient descent along with $\alpha$ and $\beta$ which are hyperparams for smoothing
     29     - PID Control
     30         - Cross track error
     31             - Lateral distance between reference trajectory and the vehicle
     32         - We want to minimize cross track error
     33             - This often overshoots though
     34                 - To achieve marginal stability we then use PD control
     35         - PD Control
     36             - When we are reducing error, we counter-steer to stop overshoot.
     37         - Systematic Bias
     38             - These are biases in our system that should be accounted for to stop oscillation
     39                 - Like tire alignment
     40         - PID
     41             - P = Proportional
     42             - I = Integral (solves for bias term)
     43             - D = Differential (solves oscillation without considering bias)
     44                 - These are the three parts of the equation for control
     45         - Control gains
     46             - These are the hyper-params for the PID
     47                 - Twiddle can solve this (coordinate descent)
     48                     - We change the hyperparams individually, grading each, and updating bumping factors
     49 - Videos 323 - 363
     50     - SLAM
     51         - Simultaneous Localization and Mapping
     52             - Localization is assuming we have a map