2. Outline
I. Overview of MILP
II. Existing UAV path planning Implementations
III. Our Implementation
a. Common Constraints
b. Significant Variables
c. Improvements
IV. Results
V. Future Improvements
14. Physical Dynamics Constraint
∀ p∈ [ 1... N ] ∀ i∈ [ 0 ... T −1 ]
si1 p= A p sip B p f ip
The next state must always be equal to force
and velocity vectors added to current state
26. Results
Realtime Flight Simulated
using ROS
Number of UAVs: 3
Number of Waypoints per UAV:
Unlimited Random (only
achieved shown)
Field Size: 1000m x 1000m
(Flight Map scaled to 10m/unit)
Flight Time: 300s
UAV Speed: 11.76 m/s
Mean Iteration Computation Time:
0.0573s
Distance actual/distance
minimum: 1.046
28. Results
Realtime Flight Simulated
using ROS
Number of UAVs: 6
Number of Waypoints per UAV::
Unlimited Random (only
achieved shown)
Field Size: 1000m x 1000m
(Flight Map scaled to 10m/unit)
Flight Time: 300s
UAV Speed: 11.76 m/s
Mean Iteration Computation Time:
0.107s
Distance actual/distance
minimum: 1.063
32. Conclusions
● MILP implemented through RHC with
Connected Components works very well
● The interactions between the ROS
framework and the MILP solver become
unstable with large amounts of UAVs
● There is a lack of concurrency between
the information from ROS and the found
solutions
33. Possible Future Research
● Improve relations between MILP and ROS
● Dynamic RHC
● Smart Connected Components