Optitrack on Raspberry Pi: Difference between revisions

From Rsewiki
 
(17 intermediate revisions by the same user not shown)
Line 40: Line 40:
==Install ASTA version ==
==Install ASTA version ==


The Asta version is adapted to the Asta optitrack version and includes data extraction for solid bodies, and should be rather easy to include in other projects.
The Asta version is adapted to the Asta Optitrack version and includes data extraction for rigid bodies only.
 
This is not ROS based. If you use ROS, then you should use the ROS package instead.
This is not ROS based. If you use ROS, then you should use the ROS package instead.


Line 58: Line 59:
  cmake ..
  cmake ..
  make
  make
==Install ASTA drone_ctrl version ==
This version includes an interface to drone_ctrl flight controller, Raspberry camera, Optitrack and GNSS.
Some additional packages are needed,
* see http://rsewiki.elektro.dtu.dk/index.php/Linux_tools for general tools
* see http://rsewiki.elektro.dtu.dk/index.php/Raspberry_Camera_API#Raspicam for camera API support - and optionally streaming software.
cd
svn co svn://repos.gbar.dtu.dk/jcan/mobotware/drone_ctrl
The raspberry pi interface to Teensy, Optitrack and GPS is then in the directory
cd ~/drone_ctrl/trunk/drone_mission
mkdir build
cd build
cmake ..
make
./mission
It will not run a mission (out of the box), but start the interface and allow configuration using the Python GUI in
cd ~/drone_ctrl/trunk/drone_gui
python3 drone_gui.py
==== Modify local IP ====
in the main.cpp file the IP of the raspberry on the wifi connection to the optitrack access point is needed.
Find the IP
ifconfig
the wlan0 should be connected and have an IP in the area 192.168.1.XX.
If not, look at the settings described below and possibly reboot the raspberry.
Modify the ''main.cpp'' code
There is a findIP(), but it works only if the Raspberry has just one IP, and there is therefore a line with a hard coded IP, as shown here:
/**  (part of main.cpp)
  * setup parameters and start */
bool startNatNetConnection(const char * argv0)
{ // find IP/name of server and this machine
  const int MIL = 100;
  char ipStr[MIL];
  const char * optitrackServerInAsta = "DESKTOP-OVLVRUD.local";
  // IP of this machine
  findIP(ipStr, MIL);
  strncpy(ipStr, "'''192.168.1.33'''", MIL);
  // generate an argc, argv set
  const char * argv1[4] = {argv0, optitrackServerInAsta, ipStr, "\0"};
  // run NatNet connection for optitrackServer
  // also starts threads for package reception
  // takes 3 string parameters (app name, server IP/name, client IP)
  bool started = setup(3, (char **)argv1);
  return started;
}
Modify as needed and recompile the code.


It should produce an executable called optiAsta, run by:
It should produce an executable called optiAsta, run by:


  cd ~/optiAsta/build/optiAsta
  ~/optiAsta/build/optiAsta
 
==== Class description ====
 
This version includes 2 classes
 
The first holds position and orientation (in quartonions)
in the optitrack coordinate system
 
class URigid
{
public:
  int ID;
  float pos[3];  // position in m in the Opritrack coordinates
  float rotq[4]; // orientation - quartonions - I think
  float posErr;
  bool valid;    // true, if visible
  void printRigid(); // print on console
};
 
and a class to hold all rigid bodies
 
class UOptitrack
{
protected:
  // positions
  static const int MAX_MARKERS = 30;
  URigid * markers[MAX_MARKERS] = {nullptr};
  int markersCnt = 0;
  double timestamp;
public:
  void print();
  URigid * findMarker(int id);
public:
  // major NatNet unpack, modified version of the one in PacketClient.cpp
  void unpack(char * pData);
};
 
One instance of UOptitrack is created in the main.cpp file.
Use the 'findMarker(int ID)' to get a pointer to a URigid with the position of that rigid body.
 
The function 'unpack(...)' is called in a thread every time there is a new data frame from the optitrack system.


== Raspberry setup ==
== Raspberry setup ==

Latest revision as of 13:09, 7 July 2021

Using Optitrack and thin NatNet client using C++

Back to ASTA motion capture

Use

The PacketClient described here is a "Direct Depacketization" code for NatNet and is not using the NatNat SDK, and thus have to be updated if the data structure used in NatNet is changed.

The Git repository is maintained, so you should keep the modifications to this file small, so that future updates are easily integrated.

Install original sample version

Optional install, you may want to use the version adapted to Asta, see next section.

Get client from https://github.com/b4be1/packet-client-linux e.g. directly to the raspberry:

git clone https://github.com/b4be1/packet-client-linux.git
cd packet-client-linux
mkdir build
cd build
cmake ..
make

Filelist:

NatNetClient.py    - python client
requirements.txt   - almost empty
CMakeLists.txt     - cmake instructions
PacketClient.cpp   - source code
README.md          - how to use/compile

The CMakeLists.txt is rather short (only the pthread library is needed):

cmake_minimum_required(VERSION 2.8)
project(PacketClient)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -Wall")
add_executable(PacketClient PacketClient.cpp)
target_link_libraries(PacketClient pthread)

Install ASTA version

The Asta version is adapted to the Asta Optitrack version and includes data extraction for rigid bodies only.

This is not ROS based. If you use ROS, then you should use the ROS package instead.

Get the source code

Install prerequisites.

sudo apt install subversion cmake

Get from repository

cd
svn co svn://repos.gbar.dtu.dk/jcan/asta optiAsta
cd optiAsta
mkdir build
cd build
cmake ..
make

Install ASTA drone_ctrl version

This version includes an interface to drone_ctrl flight controller, Raspberry camera, Optitrack and GNSS.

Some additional packages are needed,

cd
svn co svn://repos.gbar.dtu.dk/jcan/mobotware/drone_ctrl

The raspberry pi interface to Teensy, Optitrack and GPS is then in the directory

cd ~/drone_ctrl/trunk/drone_mission
mkdir build
cd build
cmake ..
make
./mission

It will not run a mission (out of the box), but start the interface and allow configuration using the Python GUI in

cd ~/drone_ctrl/trunk/drone_gui
python3 drone_gui.py

Modify local IP

in the main.cpp file the IP of the raspberry on the wifi connection to the optitrack access point is needed.

Find the IP

ifconfig

the wlan0 should be connected and have an IP in the area 192.168.1.XX.

If not, look at the settings described below and possibly reboot the raspberry.

Modify the main.cpp code

There is a findIP(), but it works only if the Raspberry has just one IP, and there is therefore a line with a hard coded IP, as shown here:

/**  (part of main.cpp)
 * setup parameters and start */
bool startNatNetConnection(const char * argv0)
{ // find IP/name of server and this machine
 const int MIL = 100;
 char ipStr[MIL];
 const char * optitrackServerInAsta = "DESKTOP-OVLVRUD.local";
 // IP of this machine
 findIP(ipStr, MIL);
 strncpy(ipStr, "192.168.1.33", MIL);
 // generate an argc, argv set
 const char * argv1[4] = {argv0, optitrackServerInAsta, ipStr, "\0"};
 // run NatNet connection for optitrackServer
 // also starts threads for package reception
 // takes 3 string parameters (app name, server IP/name, client IP)
 bool started = setup(3, (char **)argv1);
 return started;
}

Modify as needed and recompile the code.

It should produce an executable called optiAsta, run by:

~/optiAsta/build/optiAsta

Class description

This version includes 2 classes

The first holds position and orientation (in quartonions) in the optitrack coordinate system

class URigid
{
public:
 int ID;
 float pos[3];  // position in m in the Opritrack coordinates
 float rotq[4]; // orientation - quartonions - I think
 float posErr;
 bool valid;    // true, if visible
 void printRigid(); // print on console
};

and a class to hold all rigid bodies

class UOptitrack
{
protected:
 // positions
 static const int MAX_MARKERS = 30;
 URigid * markers[MAX_MARKERS] = {nullptr};
 int markersCnt = 0;
 double timestamp;
public:
 void print();
 URigid * findMarker(int id);
public:
 // major NatNet unpack, modified version of the one in PacketClient.cpp
 void unpack(char * pData);
};

One instance of UOptitrack is created in the main.cpp file. Use the 'findMarker(int ID)' to get a pointer to a URigid with the position of that rigid body.

The function 'unpack(...)' is called in a thread every time there is a new data frame from the optitrack system.

Raspberry setup

On the Raspberry Pi set the wireless network to access the dedicated Optitrack network

Modify the /etc/wpa_supplicant/wpa_supplicant.conf by adding the network groups. The 5Ghz SSID has in this case higher priority (higher number).

ctrl_interface=DIR=/var/run/wpa_supplicant GROUP=netdev
ctrl_interface_group=0
update_config=1
country=DK

network={
       ssid="asta_optitrack"
       psk="asta2020"
       priority=1
}

network={
       ssid="asta_optitrack_5G"
       psk="asta2020"
       priority=2
}

Restart the network by reboot or just restart the network:

sudo /etc/init.d/networking restart

Optitrack setup

In

  • EDIT mode (bottom of window left)

In 'View' and 'Data streaming pane'

  • select 'local interface' to the IP address of the server (192.168.1.129) and
  • deselect other streaming services (VRPN and Trackd) if not used by others

In 'View' and 'Assets Pane'

  • check the relevant rigid bodies

Then

  • go live (bottom of window left)

and assuming there is at least one rigid body defined and visible, streaming should commence.

Modify the original source

The NatNet version should be changed - it is hard-coded in source file.

 cd ~/packet-client-linux
 nano PacketClient.cpp

A bit down change the NatNetVersion to 3.1 (to fit Motive version 2.2)

// Versioning
int NatNetVersion[4] = {3, 1, 0, 0};
int ServerVersion[4] = {0, 0, 0, 0};

The multicast address is "239.255.42.99" and should not be changed.

Compile the client

Compile

cd ~/packet-client-linux/build
make

Run the client

cd ~/packet-client-linux/build
./PacketClient DESKTOP-OVLVRUD.local 192.168.1.136

where DESKTOP-OVLVRUD is the hostname of the Optitrack server (in ASTA) and 192.168.1.136 is the IP of the raspberry pi in this case.

Data in client

The client should then print package data on the console. The frame data packet looks like this:

Begin Packet
-------
Message ID : 7
Byte count : 220
Frame # : 285290
Marker Set Count : 0
Rigid Body Count : 1
ID : 20
pos: [0.73,0.57,-0.15]
ori: [-0.01,-0.00,1.00,0.02]
Mean marker error: 0.00
Tracking Valid: True
Skeleton Count : 0
Labeled Marker Count : 4
ID  : [MarkerID: 1] [ModelID: 20]
pos : [0.72,0.50,-0.15]
size: [0.01]
err:  [0.00]
ID  : [MarkerID: 2] [ModelID: 20]
pos : [0.78,0.53,-0.14]
size: [0.01]
err:  [0.00]
ID  : [MarkerID: 3] [ModelID: 20]
pos : [0.74,0.62,-0.15]
size: [0.01]
err:  [0.00]
ID  : [MarkerID: 4] [ModelID: 20]
pos : [0.66,0.64,-0.15]
size: [0.01]
err:  [0.00]
Timestamp : 2377.417
Mid-exposure timestamp : 12335057189261
Camera data received timestamp : 12335057223494
Transmit timestamp : 12335057235053
End Packet

The data is thus available in the client and should be integrated into the intended application (by further source code editing).