Visa alla sidor

Project task description

The task to be solved in the project can be thought of as an object search task constrained by having to pass a number of gates at known positions along the way. To assist the system in terms of localization there are a number of markers in the environment. These markers are ArUco markers of known size and known pose. The objects are traffic signs printed on paper. The aim is to navigate along the track defined by the gates and find objects at the same time. 

You are allowed to start a mission at an arbitrary pose within 1m from the center of the first gate to be passed.

Task Environment

This section describes the definition of the task environment.

Map

Since the UAVs have very limited sensing obstacle avoidance is dependent on knowing the position at all times well enough and knowing where things that one can collide with area before hand. You will be give a map describing the known environment. The map is given in .json format. An example is given here.

Airspace

The airspace is defined by a bounding rectangle that is axis aligned. It is not safe to move outside the airspace as you might harm people and UAV by, for example, running into a wall. During the contest, your UAV might be disqualified when flying outside. 

Markers

The pose and ID of all ArUco markers is given by the map. The markers can be placed in any pose (inside and outside the airspace. The markers will have unique IDs, ie. a marker gives a single fully defined pose.

 

Gates

The gates corresponds to special obstacles that should be passed through to be defined as "cleared". It will typically be a "hole" with vertical walls around and below/above. You have to pass though it in the right direction at least once before the next gate to be counted as "cleared". 

To be cleared you need to have passed in the right direction from at least 0.1m before the gate to at least 0.1m after the gate.

You need to pass them in order and the run is considered over when you pass the last one, even if all objects are not found.

Each gate will have one or more markers in the near vicinity to facilitate identifying and positioning it.

Objects

Besides navigation, the main objective in the task is to find objects. These are represented by a set of Swedish traffic signs, printed in color on paper. All objects are known in advance, in terms of size and appearance, but the location is unknown. They can be placed anywhere in the environment

  • both on the floor and vertically
  • both inside and outside the airspace

When placed vertically they will be "upright" (ie have the angle they would have when found in trafic)

There can be more than one object of each kind but these will be a minimum of 4m apart. Besides identifying them they should be positioned.

 

Obstacles

In addition to gates there can be any number of other obstacles in the environment. An obstacle consists of one or several thin vertical "walls". By thin here means something on the order of cardboard. It is thin to reduce the disturbance that the height sensor gets from it.

Some obstacles will not cover the entire airspace in the vertical direction and can thus be flown under, through or over depending on configuration.

The complete track through all gates can be flown without having to pass over or under any obstacles, except gates which are 40x40 cm openings approximately 21 cm above ground. Some obstacles can be flown over or under, though never required to do so, it may be significantly faster, and some areas may be completely unreachable without it -- these “hard to get to” areas could potentially contain detectable objects.

 

 

Contest environment

There will be a contest to compere the solutions towards the end of the course. It will take place in the atrium in the E-building. You are very much encouraged to visit this area and before the contest so that you get a feeling for how your system behaves there.

 

 

Ranking of solutions

In the context of contests, challenges, etc different solutions will be compared. For the full project task, the ranking is defined by

  • G=number of cleared gates
  • O=number objects correctly identified and positioned with a meter (we deduct 0.25 points for each detection outside, and extraneous detections inside are ignored. If there are no acceptable detections, the score is set to -10 points.)
  • T=time to the end of the run (system stops, passing the last gate, crash, out of battery, etc)

according to

  1. Min(G,O)
  2. O
  3. T

The rationale is that we want to emphasise system integration and thus you are primarily ranked by the weakest of the navigation and perception abilities, encouraging building a uniformly system rather than a system that excels at one thing.

 

Minimum functional requirements for passing

The minimum functional requirements on the system for passing the project is that you implement a single integrated system that can

  • Compensate for the drift inherent in the dead-reckoning system of the UAV so that the distance it can move without getting lost, "on a good day", is only be limited by the battery life.
  • Autonomously perform path planning and path execution such that the UAV can, "on a good day", make a plan and fly through a previously unknown course and clear at least 4 gates (minimum distance 1m part and involving accumulated turns of at least 270 degrees).
  • Detect at least one object whose location is not known by the system and is positioned after moving at least 4m from the start position of the UAV. 

By "on a good day" is meant that it does not have to work every time and it is does not need to be robust.

It should be clear from the way the system is demonstrated that a successful demonstration is not the result of "hard coding" but rather a general solution.