Abstract

While Ocado has many challenging problems in hardware, robotics, distributed systems, iPhone/Android, web technologies and so on, in this talk I’ll describe and formulate a handful of the optimisation and modelling problems we deal with, and their similarities and differences to the archetypal computer science problems. These range across:

• optimising our delivery routes (a kind of TSP or VRP )

• packing our customers orders (a kind of 3d bin-packing)

• calculating optimal driving paths (Dijkstra shortest path)

• forecasting product demand (machine learning)

• simulating/optimising our warehouse (a discrete event simulation with many different embedded optimisation sub-systems)

• optimising our website product recommendations • helping customers shop more quickly by suggesting predictions of what they will buy

Ocado’s scale (1 million items picked per day, 100000 orders delivered per week) ensures that in all of these areas, optimising to attain the last fraction of a percent of improvement is of genuine value, and not just a theoretical nicety. However all of these problems have additional complications and constraints which must somehow be included in the theoretical formulation.

BIO

Dr Darley started his career with a Maths degree from Trinity College, Cambridge, followed by a move to Harvard for a Masters in Computer Science and a PhD (jointly between Applied Maths, Computer Science and Economics), including a spell at the Santa Fe Institute. His research studied both the dynamics of groups of “economic agents” when traditional assumptions of perfect rationality, complete information, etc. are relaxed, and phase transitions in problem difficulty of NP-complete problems. He then spent a few years working at Bios Group in Santa Fe, modelling and optimising for organisations such as Unilever, P&G, Ford, Nasdaq. He wrote a book summarising several years of research on the Nasdaq stock market, including accurate predictions of changes to market dynamics which would come to be caused by “decimalisation”. In 2000 he moved back to the UK to set up his own software and consulting business which provided sophisticated modelling and optimisation to European businesses. Since early 2009 he has been chief scientist with Ocado.