Autonomous smart traffic control is proposed to relieve traffic congestion for bus scheduling, to intelligently accomplish tasks such as on-demand dynamic passenger pickup or drop-off.
What Could Cause A VW Tiguan's Radiator Fan To Stop Working
Vtc9252019
1. Brain Storm Optimized Swarm
Collaboration for Bus Scheduling
Liangxi Liu, Siqing Ma, Jun Steed Huang
Southern University of Science and Technology, Shenzhen 518000, China
Cultivation of Guangdong college students' scientific and technological innovation.
(``Climbing Program'' special funds. No. pdjhb0445.)
September 22-25, 2019, Honolulu, USA
2019 IEEE 90th Vehicular Technology Conference
2. About the Authors
Liangxi Liu, Research Assistant
Siqing Ma, Teaching Assistant
Jun Steed Huang, Research Fellow
Co-operating agency:
4. In order to provide the design guidance for a swarm of unmanned buses
which supposed to intelligently accomplish tasks such as on-demand
dynamic passenger pick-up or drop-off.
To analyze this problem, we provides an artificial sparrow model of
neighborhood collaboration strategy. The brain storm optimization (BSO) is
further applied to enable the global path planning of the bus swarm.
Introduction
5. Dynamic Scheduling Framework
Low-level path planning —— Modified Dynamic Window Approach
The key idea of the DWA* is tantamount to construct a feasible velocity
interval and choose a velocity at which to optimize the possible motion path.
It allows the individual to reach the destination with the best orientation,
leaving the best safety collision distance, at the fastest speed.
Objective function:
* https://en.wikipedia.org/wiki/Dynamic_window_approach
6. The Finsler geometric measure
Traditional DWA did not consider the collaboration problem when applied in
a swarm of individuals.
In Finsler manifold*, the measured coverage area will be smaller or larger
than the Euclidean measured counterpart, which can used to avoid the
vehicle getting too close or too far from each other.
https://en.wikipedia.org/wiki/Finsler_manifold
7. High-level scheduling decision making
The high-level relationship variable between the vehicle and the destination,
determines which destination the vehicle goes.
A many-to-one relationship
Goal refresh strategy
8. Optimization framework formation
A basic objective function should consider the following three aspects:
1) the distance between the vehicle and the target point;
2) the demand of the destination;
3) the result of the previous round of scheduling.
Where the user demand can be obtained as:
Objective function:
9. Brain storm optimization (BSO) algorithm is a new and promising swarm
intelligence algorithm, which was proposed by Yuhui Shi in 2011*, it simulates the
human brainstorming process.
Brain storm optimization algorithm imitates the human brainstorming process.
Through the convergent operation and divergent operation, individuals in BSO are
grouped and diverged in the search space.
BSO is further applied to enable the dynamic optimization and adjustment of bus
swarm (High-level Decision making).
Brain Storm Optimization
* https://en.wikipedia.org/wiki/Brain_storm_optimization_algorithm
10. Flow chart
BSO is an optimization algorithm that was
inspired by human being creative problem
solving process. This new algorithm combines
the advantages of swarm intelligence and data
mining, choosing better solution based on data
analysis.
Each solution in this swarm intelligence
optimization algorithm is regarded as a data
point, and the optimal solution is found by
clustering data points.
11. Algorithm detail
The solution update to their new solution by:
Xnew
d
= Xselected
d
+ ξ t × N μ, σ
where Xselected
d
is the d-th dimension of the individual selected to generate new
individual; Xnew
d
is the d-th dimension of the individual newly generated; N(μ,
σ) is the Gaussian random function with mean μ and variance σ; the ξ(t) is a
coefficient that weights the contribution of the Gaussian random value.
For the simplicity and for the purpose offline tuning, the ξ can be calculated as
ξ t = logsig(
0.5×T–t
𝑘
) × rand()
.
where logsig() is a logarithmic sigmoid transfer function, T is the maximum
number of iterations, and t is the current iteration number, k is for changing
logsig() function’s slope, and rand() is a random value within (0,1). The transfer
function formula logsig() is defined as follows:
logsig a =
1
1+exp(–𝑎)
12. Simulation
The original Player/Stage simulator* is chosen in this section, which is a robot
simulation platform and supports simulating a large number of individuals.
The simulated scenario.
* https://en.wikipedia.org/wiki/Player_Project
13. Results
During the simulation, the weight of the target points is changing from {3, 3, 3, 3, 3}
to {1, 2, 1, 3, 3}, which simulates the dynamics of passenger demands.
14. Conclusion
--The collaborative control strategy based on the Finsler geometric
measure is simple and easy to implement, which have an apparently
controlling effect on swarm distribution and good for preventing
collisions.
--The BSO algorithm use random solutions to replace clustering
centers in solution set, which can help BSO to avoid achieving a
premature convergence of local optimal solution.
15. Acknowledgement
http://www.ieeevtc.org/vtc2019fall/
--Thanks go to Prof. Yuhui Shi
(https://en.wikipedia.org/wiki/Yuhui_Shi)
for providing guidance on using of his BSO algorithm
--Thanks go to Prof. Oliver Yang
(http://www.site.uottawa.ca/~yang/)
for providing guidance on organizing the vehicle research paper.