Remove last mentions of nuserdata from codebase, improve README, ad…

…d reference to MIT talk.

PiperOrigin-RevId: 495488238
Change-Id: Ic9f2a7a1d80025f8936a84fa3d9e883ed65f761c

README.md

@@ -15,31 +15,37 @@
 
 MJPC allows the user to easily author and solve complex robotics tasks, and currently supports three shooting-based planners: derivative-based iLQG and Gradient Descent, and a simple yet very competitive derivative-free method called Predictive Sampling.
 
+- [Overview](#overview)
 - [Installation](#installation)
-- [Getting Started](#getting-started)
 - [Predictive Control](#predictive-control)
 - [Contributing](#contributing)
 - [Known Issues](#known-issues)
 - [Citation](#citation)
 - [Acknowledgments](#acknowledgments)
 - [License and Disclaimer](#license-and-disclaimer)
 
-## Installation
+## Overview
 
-You will need [CMake](https://cmake.org/) and a working C++20 compiler to build MJPC. We recommend using [VSCode](https://code.visualstudio.com/) and 2 of its extensions ([CMake Tools](https://marketplace.visualstudio.com/items?itemName=ms-vscode.cmake-tools) and [C/C++](https://marketplace.visualstudio.com/items?itemName=ms-vscode.cpptools)) to simplify the build process.
+To read the paper describing this software package, please see out [preprint](https://arxiv.org/abs/2212.00541).
 
-1. Clone the repository: `git clone https://github.com/deepmind/mujoco_mpc.git`
-2. Configure the project with CMake (a pop-up should appear in VSCode)
-3. Build and run the `mjpc` target. This will open and run the graphical user interface.
+For a quick video overview of MJPC, click below.
 
-## Getting Started
+[![Video](http://img.youtube.com/vi/Bdx7DuAMB6o/hqdefault.jpg)](https://dpmd.ai/mjpc)
 
-For a video overview of MJPC, click below.
+For a longer talk at the MIT Robotics Seminar describing our results, click below.
 
-[![Getting Started](http://img.youtube.com/vi/Bdx7DuAMB6o/hqdefault.jpg)](https://dpmd.ai/mjpc)
+[![Talk](http://img.youtube.com/vi/2xVN-qY78P4/hqdefault.jpg)](https://www.youtube.com/watch?v=2xVN-qY78P4)
 
 For a detailed dive of the graphical user interface, see the [MJPC GUI](docs/GUI.md) documentation.
 
+## Installation
+
+You will need [CMake](https://cmake.org/) and a working C++20 compiler to build MJPC. We recommend using [VSCode](https://code.visualstudio.com/) and 2 of its extensions ([CMake Tools](https://marketplace.visualstudio.com/items?itemName=ms-vscode.cmake-tools) and [C/C++](https://marketplace.visualstudio.com/items?itemName=ms-vscode.cpptools)) to simplify the build process.
+
+1. Clone the repository: `git clone https://github.com/deepmind/mujoco_mpc.git`
+2. Configure the project with CMake (a pop-up should appear in VSCode)
+3. Build and run the `mjpc` target in "release" mode (VSCode defaults to "debug"). This will open and run the graphical user interface.
+
 ## Predictive Control
 
 See the [Predictive Control](docs/OVERVIEW.md) documentation for more information.

docs/OVERVIEW.md

@@ -149,8 +149,6 @@ It is also possible to create GUI elements for  parameters that are passed to th
 - `lower_bound`: Real specifying lower bound of GUI slider.
 - `upper_bound`: Real specifying upper bound of GUI slider.
 
-Finally, the task file must specify `<size nuserdata="1"/>`, as that memory space is used by the planners.  In our task examples, this line appears in the file [`common.xml`](../mjpc/tasks/common.xml) that also shares visual definitions across the tasks.
-
 ### Residual Specification
 
 As mentioned above, the cost is a sum of terms, each computed as a (scalar) norm of a (vector) residual. Each term is defined as a [user sensor](https://mujoco.readthedocs.io/en/latest/XMLreference.html#sensor-user). The sensor values constitute the residual vector (implemented by the residual function, see below). The norm for each term is defined by the [`user` attribute](https://mujoco.readthedocs.io/en/latest/modeling.html#cuser) of the respective user sensor, according to the following format:

mjpc/agent.cc

@@ -44,9 +44,6 @@ void Agent::Initialize(mjModel* model, mjData* data,
                        ResidualFunction* residual,
                        TransitionFunction* transition) {
   // ----- model ----- //
-  if (!(model->nuserdata == 1)) {
-    mju_error("Model passed to Agent requires nuserdata == 1\n");
-  }
   if (this->model_) mj_deleteModel(this->model_);
   this->model_ = mj_copyModel(nullptr, model);  // agent's copy of model
 

mjpc/tasks/common.xml

@@ -1,5 +1,4 @@
 <mujoco>
-  <size nuserdata="1"/>
 
   <visual>
     <headlight ambient=".4 .4 .4" diffuse=".8 .8 .8" specular="0.1 0.1 0.1"/>

mjpc/test/testdata/cartpole_task.xml

@@ -1,7 +1,7 @@
 <mujoco model="Cart-Pole Swing-Up">
   <include file="cartpole.xml" />
 
-  <size nuserdata="1" memory="40K"/>
+  <size memory="40K"/>
 
   <custom>
     <numeric name="agent_planner" data="0" />

mjpc/test/testdata/particle_task.xml

@@ -1,7 +1,7 @@
 <mujoco model="Particle Control">
   <include file="particle.xml" />
 
-  <size nuserdata="1" memory="10K"/>
+  <size memory="10K"/>
 
   <custom>
     <numeric name="task_risk" data="1.0" />