MPPI

README.md

@@ -18,7 +18,7 @@ Google DeepMind.
 MJPC allows the user to easily author and solve complex robotics tasks, and
 currently supports multiple shooting-based planners. Derivative-based methods include iLQG and
 Gradient Descent, while derivative-free methods include a simple yet very competitive planner
-called Predictive Sampling and the Cross Entropy Method (with diagonal covariance).
+called Predictive Sampling, the Cross Entropy Method (with diagonal covariance), and Model Predictive Path Integral control (with fixed, diagonal covariance).
 
 - [Overview](#overview)
 - [Graphical User Interface](#graphical-user-interface)

docs/OVERVIEW.md

@@ -289,6 +289,9 @@ This library includes multiple planners that use different techniques to perform
 - **Cross Entropy Method**
   - all properties of Predictive Sampling
   - refits a nominal policy to mean of elite samples instead of using the best
+- **Model Predictive Path Integral Control**
+  - all properties of Predictive Sampling
+  - refits a nominal policy to a weighted average of perturbed policies centered around the previous nominal policy
 - **Gradient Descent**
   - requires gradients
   - spline representation for controls

mjpc/CMakeLists.txt

@@ -79,6 +79,8 @@ add_library(
   planners/model_derivatives.h
   planners/cross_entropy/planner.cc
   planners/cross_entropy/planner.h
+  planners/mppi/planner.cc
+  planners/mppi/planner.h
   planners/robust/robust_planner.cc
   planners/robust/robust_planner.h
   planners/sampling/planner.cc

mjpc/planners/include.cc

@@ -21,6 +21,7 @@
 #include "mjpc/planners/gradient/planner.h"
 #include "mjpc/planners/ilqg/planner.h"
 #include "mjpc/planners/ilqs/planner.h"
+#include "mjpc/planners/mppi/planner.h"
 #include "mjpc/planners/planner.h"
 #include "mjpc/planners/robust/robust_planner.h"
 #include "mjpc/planners/sampling/planner.h"
@@ -32,7 +33,8 @@ const char kPlannerNames[] =
     "iLQG\n"
     "iLQS\n"
     "Robust Sampling\n"
-    "Cross Entropy";
+    "Cross Entropy\n"
+    "MPPI";
 
 // load all available planners
 std::vector<std::unique_ptr<mjpc::Planner>> LoadPlanners() {
@@ -46,6 +48,7 @@ std::vector<std::unique_ptr<mjpc::Planner>> LoadPlanners() {
   planners.emplace_back(
       new RobustPlanner(std::make_unique<mjpc::SamplingPlanner>()));
   planners.emplace_back(new mjpc::CrossEntropyPlanner);
+  planners.emplace_back(new mjpc::MPPIPlanner);
   return planners;
 }