As I mentioned last year, Andrea Tassi and I have been working on applications of mmWave in vehicular communications. Andrea has been visiting Trung Duong at Queen’s University in Belfast and has presented a talk on this work. You can find the slides at Andrea’s website.
In two weeks I will be attending the STM2016 workshop in Tokyo on spatial-temporal modeling. During the workshop I will be presenting some work with Nourddine Azzaoui and Gareth Peters on the simulation of a general class of non-stationary -stable processes. In this post, I want to provide some background to this work.
Update 2/11/2016: At present, there is new development of the testbed. If you are interested in specific features, you should contact Michal Čap.
A key challenge in understanding market-based on-demand transport services is that it is not just a local problem; i.e., the behavior of an individual market. Since there are usually a number of allocations going on at the same time and individual markets are changing over time another approach is required. In fact, a rigorous evaluation requires a study of how the on-demand transport interacts with the underlying transportation network and even other dynamics of the city as we alluded to in our working paper here.
The field of multiagent systems provides a good framework to study how markets behave when they are embedded in transportation networks. Within this framework, drivers, passengers and even providers are modeled as autonomous agents. This means that each driver, passenger and provider makes its own decisions based on their own preferences and the information that is available to them. Using the multiagent systems framework allows us to capture differences in the preferences of individual drivers, passengers and providers and also competition between them.
In recent papers (see here and our preprint here), we studied how the mechanism performed on in the Hague with a realistic demand profile for passenger requests. The basis of our network-scale evaluation was a simulation tool: the mobility services testbed. This simulation tool was developed by Michal Certicky and Michal Jakob at the Czech Technical University in Prague.
The mobility services testbed provides an easy way of implementing different market mechanisms in the context of on-demand transport. You can get it from github here.
For an overview of other aspects of mechanisms for on-demand transport, see these posts:
If you are a company providing or a municipality supporting on-demand transport services, there is an important decision that you must make: what should be the structure of the service? By structure, I mean how drivers and passengers can interact, or how payments are made. For instance, traditional taxi services differ in their structure from providers such as Uber.
In an earlier working paper (available here) and the post here, I discussed the differences between the main classes of on-demand transport services: hackney carriage; dispatcher; dial-a-ride; and market-based. However, the classification was limited in that it did not distinguish easily between services within a given class.
With the rise in market-based on-demand transport services (e.g., Uber), there is now a need to understand the different ways that these services can operate. And this is what I intend to describe here.
Recently, I have had some reason to reflect on the perspective that I adopt within my research. Here, I want to explore these thoughts further.
A one-line summary of my approach could be: “local to global and back again”. The sub-text would then read: “… and implications for system design”. Of course, these could possibly be some of the most overloaded words in (technical) English literature, so it is worth elaborating.
In July, the International Symposium on Information Theory will be held in Barcelona, Spain. One of the papers that will be appearing there is some recent work I have done with Mauro de Freitas, Laurent Clavier, Alban Goupil, Gareth Peters, and Nourddine Azzaoui. We have been considering variations on the additive -stable noise channel , where is an -stable random vector.
These kinds of channels appear in various communication systems including wireless and quite recently molecular. As such, it is interesting to try and compute the capacity of these channels. The special case with a power constraint has been extensively studied (it is the Gaussian case!), but in general the capacity is not well understood for other values of with any constraints.
Enter our paper. We considered the case where the noise is an isotropic -stable random vector. So, the channel is the additive isotropic -stable noise () channel. Our results? A quick summary:
- The optimal input for the channel subject to a constraint exists and is unique.
- The capacity subject to , is lower bounded by a function of the form: . ( is just the minimum of the elements of and is a constant that depends on the noise parameters)
We also had a brief look at the extension to parallel channels, but for that you will need to read the paper.
For the official summary…
Title: Achievable rates for additive isotropic alpha-stable noise channels
Abstract: Impulsive noise arises in many communication systems—ranging from wireless to molecular—and is often modeled via the -stable distribution. In this paper, we investigate properties of the capacity of complex isotropic -stable noise channels, which can arise in the context of wireless cellular communications and are not well understood at present. In particular, we derive a tractable lower bound, as well as prove existence and uniqueness of the optimal input distribution. We then apply our lower bound to study the case of parallel -stable noise channels and derive a bound that provides insight into the effect of the tail index on the achievable rate.
During 2014 – 2015, one of the key research problems I was working on was how to understand and design on-demand transport mechanisms. On-demand transport is about how an individual or group can get from point A to point B at a time of their choosing. Common examples are taxi services and now new services such as Uber.
Working in a computer science department, primarily with collaborators Michal Jakob and Nir Oren, I was concerned with how passenger journeys are allocated to drivers and how each journey was priced, together called a mechanism, from an algorithmic perspective. I.e., how can these allocations and pricing be done in a computationally efficient way so that passengers get to where they want to go on time, each driver can earn a living, and service providers (e.g., Uber) can make a profit.
The problem of choosing a mechanism is known in economics as the mechanism selection problem, and must account for a range of technical, social and financial issues. For example, can the available computing resources compute the allocation and pricing quickly enough? Or, are passengers or drivers prepared to bid for a journey (a key problem for auction-based approaches)?
We have explored the problem of mechanism selection in on-demand transport by first enumerating the possible mechanisms and evaluating their performance. We observed in this working paper that each approach (e.g., hackney carriages, taxi dispatcher models, and Uber-type approaches) can be differentiated by limitations on communication and financial exchanges.
After enumerating the possibilities, we have begun to explore how the different mechanisms perform in terms of metrics such as the proportion of passengers served, costs of journeys and provider profit. In particular, we have published work in:
(1) Malcolm Egan, Martin Schaefer, Michal Jakob, and Nir Oren, “A double auction mechanism for on-demand transport networks”, in the Proc. PRIMA 2015: Principles and Practice of Multi-Agent Systems, (2015).
(2) Malcolm Egan, and Michal Jakob, “A profit-aware negotiation mechanism for on-demand transport services”, in the Proc. of the European Conference on Artificial Intelligence (ECAI), (2014)
(3) Malcolm Egan and Michal Jakob, “Market mechanism design for profitable on-demand transport services”, accepted for Transportation Research Part B: Methodological.
A key feature of our article (3) is that we provide and justify an agent-based model for on-demand transport services that captures the preferences of passengers. This means that we do not assume that every passenger will accept whatever they are offered, which is commonly assumed in previous work on on-demand transport mechanisms.
The next step is to continue to study the mechanism selection problem by understanding the requirements and performance of other on-demand transport mechanisms. At the end of the day, we hope that this work will aid new providers and municipalities to decide the kinds of mechanisms they want to support to match the unique economic, social and technical features of their cities. We believe this will provide a means to meet the needs of passengers, drivers, and providers in each city in a sustainable way.
To read more, see the next posts in this series:
A quick note that I have finished up in Prague and I have begun another postdoc in Université Blaise Pascal in Clermont Ferrand, France. In between I spent a month working with Andrea Tassi at the University of Bristol. More details of this collaboration to come.
Over the last two weeks I have been in Japan visiting the Institute for Statistical Mathematics. During the first week I attended the STM and CSM workshops organized by Prof. Tomoko Matsui and Dr. Gareth Peters. I gave two talks and slides/videos are available here:
I’m in London at the moment, visiting Gareth Peters at UCL to think about alpha-stable random variables and their applications. Tomorrow, I’m also presenting in the Agents and Intelligent Systems Seminar at Kings College London (see the link for location and time). The details are:
Title: From Taxis to Uber: Market Design for On-Demand Transport Services
Abstract: Uber is one of several recent companies adopting a business model that lies in stark contrast with the standard approach used by taxi services–evidenced by the highly publicized legal difficulties. Underlying Uber’s business model is a new architecture, which governs how commuters, drivers, and the company interact with each other. In order to understand key properties of this new architecture, we introduce an agent-based model and propose a market mechanism that routes, schedules, and prices commuters, while also selecting and paying drivers. We analyze the mechanism and demonstrate the effect of varying passenger types and side information available to the company affects the profitability of the service. We then compare the performance of our approach with a mechanism based on the standard taxi architecture via simulations using realistic demand and location data from Prague, Czech Republic.