Presentation on theme: "Future Dispatching Bill Cumpston and Jason Lawrie."— Presentation transcript:
Future Dispatching Bill Cumpston and Jason Lawrie
REQUIREMENTS TAXIS As many bookings as possible Wait times not major consideration Driver not normally relevant HIRE CARS, WATS Control numbers of bookings Wait times critical for hire cars, important for WATs. Driver often relevant
Various algorithms – zone, distance, cover Fleets have different requirements Explanations are complicated Not yet perfect Dispatch is Complicated
Pickup Addresses Destination Addresses Requested Pickup Time ASAP or Pre-Booking Requested Vehicle Attributes (WAT, etc.) Our Current Algorithm Inputs
Preferred vehicle or driver Current passenger wait time Relative priority of attributes (must do maxi) Driver plotting Vehicle vacant time Blacklist preferences Our Current Algorithm Inputs
Driving time from current position Time criticality (e.g. meeting a train) Current driver earnings per hour Driver rewards earned Driver penalties incurred It will not be getting simpler!
Driver end of shift time and location Distribution (E.g. trip run fairness) Pre-allocation (E.g. private jobs) Distribution to sub-networks or friends Changes for peak or normal periods And thats not all ….
Customer experience vs. Cost Reduction? Driver fairness vs. rewards and penalties? Many things are Trade-offs
High level categories provide a guide This is combined with an input weighting Add reward or penalty scores Car or job with highest score wins Weighting Based Algorithm
Can we eliminate zones? Perhaps have some special regions (ranks) All distance calculations based on actual directions (including current time of day) Who Likes Zones and Layering?
Their Score Will increase over time until they get a job Rewards or penalties will affect their scores In a localised region (e.g. rank) the highest score in a similar vehicle will get job first What will drivers see?
Click Explain on any offer Get scores for every vehicle for that offer Can explain to drivers if needed Can be used to tweak weighting But why did.….. get job …… ?