Vehicle breakdown

1
Context-Aware Logistic Routing and
Scheduling Adriaan ter Mors, Almende Jonne Zutt,
Cees Witteveen, Algorithmics, Faculty of EEMCS,
TU Delft
Problem description
Methods

• Set of autonomous vehicles
• Transportation infrastructure of resources
• Set of transportation requestsfor each request,
  find a conflict-free shortest path
• head-on conflicts
• catching-up conflicts
• Taking into account malfunctioning vehicles and
  resources

• Traffic rules and zone control (TR)
• Only allow actions that can never lead to a
  conflict
• Fixed path scheduling (FPS)
• Insert waiting times for best schedule along
  pre-determined path
• Job Shop Scheduling with blocking
• Time window graph routing (TWGR)
• Search through graph of free time windows
• Avoiding planned movements of others

Time window graph routing

• Smart conflict prevention from explicit
  constraintchecking to off-line encoding of
  constraints into freetime windows
• Complexity from O(n2v4) to O(nv log(nv) n2v)
• Better spread of agents over space and time
  resultsin better performance
• A-search through graph of free time windows

Gridlock
Incidents
Coping with incidents

• Vehicle breakdown
• Change in transportation requests
• Arrival of new transportation requests
• Disruption on infrastructure
• impact of incident
• duration of incident

• Introduce slack into routing schedules
• Multi-objective routing
• efficiency
• reliability
• Robustness of planning and execution method
  combined
• Impact of bi-directional resources

Example time window graph
Example time window graph

• Factory shop floor
• Transportation order from one machine to another
• bi-directional lanes
• dynamic environment
• AGV container terminal
• Highly structured infrastructure
• Transportation orders in large batches
• Airport taxiing
• Congestion in peak hours and exceptional weather
  conditions
• Multi-stage routing problem
• Mostly uni-directional taxiways

The red truck has reserved its plan. Now, the
blue truck wants to plan a routeto its
destination e4.
5
Dest.
4
From a2 two free time windows in b2 can be
reached. Only the later window can reach a free
time window in c2.
3
2
The shortest-time path is to go via a5.
1
a
b
c
d
e
TWGR versus FPS
Fixed Path Scheduling is very fast, but Time
Window Graph Routing also finds a solution within
0.5s.
TWGR with heuristic TWGR no heuristic TWGR no
cycles FPS
Total vs. single plan quality
TWGR is an optimal planning algorithm for a
single-agent. Order in which agents plan is
important for individual plan quality. For the
Schiphol airport experiments, the order in which
agents plan is not of great importance to system
performance.
TWGR with heuristic TWGR no heuristic TWGR no
cycles FPS
Repeated use of Fixed Path algorithms leads to
overuse of key resources. Time Window Graph
routing provides spread of agents over space and
time.
TWGR with heuristic TWGR no heuristic TWGR no
cycles FPS
TWGR with heuristic TWGR no heuristic TWGR no
cycles FPS
Adriaan ter Mors Jonne Zutt Cees Witteveen
adriaan_at_almende.org j.zutt_at_tudelft.nl c.witteveen_at_
tudelft.nl