3
$\begingroup$

I am training Turtlebot 2 to move around the office. I come from robotics hardware background but relatively new to computer vision and robotic mapping.

Here are my two fundamental questions related to localization and mapping.

  1. The algorithm we are using to mapping is SLAM. This makes sense but my question is in general, if we have performed SLAM, then the map is created and this map can be saved for later. So is it right to say, next time if the bot moves in the same environment, it is not using SLAM but instead only localisation (since the map already exist). Is SLAM in general used only for first time mapping within an environment?

  2. If there is no mapping at all, can bots be trained to move randomly and turn back or around only if there is an obstacle or wall. This wouldn't require mapping, right? So is it right to say, bots can be moved either by mapping or without mapping, depending on the need.

$\endgroup$
2
$\begingroup$

Is SLAM in general used only for first time mapping within an environment?

Yes, SLAM is used only once to build map of the environment before navigation operation. Also, In case of significant change in the environment.

If there is no mapping at all, can bots be trained to move randomly and turn back or around only if there is an obstacle or wall. This wouldn't require mapping, right? So is it right to say, bots can be moved either by mapping or without mapping, depending on the need.

Maps are required for

  1. Localization (to know the current location of Robot). One cam localize robot without maps using visual tag or some other markers to make robot identify it location.
  2. Path planning (given destination, returns obstacle free efficient path).

Without map, it is possible to move the robot randomly and avoid obstacles. If you are using laser, you don't need learning, its easy to avoid obstacles based on sensor data.

| improve this answer | |
$\endgroup$
2
$\begingroup$
  1. SLAM is a recursive algorithm with no base case. I suppose you could have a termination criteria for the algorithm stating that "This Map is Sufficient", and then change your algorithm to localization. If you already have a map of your environment, then no, SLAM is not necessary. So to answer your first question plainly, yes.

  2. The type of agent you are referring to is a simple reflex agent without an internal model (does not track the environment state). This is one of the simplest types of agents and while yes, it can be used, it cannot make any guarantees of successful navigation.

| improve this answer | |
$\endgroup$
1
$\begingroup$

Is SLAM in general used only for first time mapping within an environment?

Mostly, yes however there have been some works where the researchers tried to update the map. You can reuse the map later if the map is static. In most cases, this is not true in real life.

If there is no mapping at all, can bots be trained to move randomly and turn back or around only if there is an obstacle or wall. This wouldn't require mapping, right? So is it right to say, bots can be moved either by mapping or without mapping, depending on the need.

As @nayab has said, you can operate in the environment without a map, however, using a map allows you to localize the robot in the map and plan the trajectory. You can still localize a robot without a map, using a global localization system (GPS/GNSS) or if it were indoor using beacons.

You can plan a trajectory using a topological map, not necessarily a metric map.

| improve this answer | |
$\endgroup$
  • 1
    $\begingroup$ "This is not true in real life." This depends on the application. $\endgroup$ – koverman47 Feb 24 at 22:04
  • $\begingroup$ I edited it is most cases. human populated environments are rarely static $\endgroup$ – Alex Feb 24 at 22:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.