In the first lesson,
You saw some of the best planning algorithms that solve motion planning problems. But we never formally defined the problem. I want to do it now.
Now, there's a word you'll probably come across a lot if you start reading material on motion planning algorithms.
The term is defined as "configuration space". All possible robot configurations in a given world. Consider the maze words you saw in the first lesson, which are all 2D grids,
The robot configuration is sometimes two, when we present it as x,y points, and sometimes the 3D space when the robot is heading.
In fact, the configuration space for the vehicle can be, depending on the motion planning algorithm we decide to use, even larger.
With this idea of configuration space in mind, we can define the motion planning problem as follows.
We have three things.
initial configuration, target configuration
and some restrictions describing how vehicles are allowed to move,
Its dynamic and environmental description.
Here, it's important to understand how this connects to other decision modules you've recently learned about.
Typically, the launch configuration is the current configuration we provide to our car.
Localized value and sensors that give us information about the car's location, speed, acceleration, etc.
The behavior layer gives us the final configuration we need, and possibly where and at what speed.
Finally, prediction completes the problem by giving us information about how the barrier area evolves in time.
In this way, the sequence of actions we generate takes into account the behavior of other vehicles and pedestrians.
If we are using more complex prediction modules, how our actions affect them.
The motion planning problem can then be solved, defined as the final sequence of actions.
The configuration space from which the robot moves, starting configuration to final configuration without encountering any obstacles. Now, I will ask you to consider configuration space in the following questions.