visualization, database systems, knowledge representation, and planning. The first part of the workshop
included invited talks from experts across these
fields. Juliana Freire at New York University spoke
about how big data techniques can be married with
interactive visualization and analysis to enable users
to find patterns in urban activity such as taxi data.
Manuela Veloso at Carnegie Mellon University spoke
on optimizing traffic flow with artificial intelligence
techniques in cities and robot localization (SLAM) in
urban environments. Craig Knoblock at the University of Southern California spoke on how differing
ontologies can impede cross-city data analysis and
how to automate the detection and merging of public data-set ontologies. Adi Botea at IBM discussed
edging risk in journey planning with a risk-averse
multimodal journey advisor that performs journey
planning, journey monitoring, and replanning.
Finally, Mike Flowers at Enigma and New York
Unviersity discussed data science in New York City,
including fire-risk models, analyzing NYC business
locations and implementing DataBridge.
The workshop concluded with contributions in
the areas of building energy efficiency, predictive
policing in New York city, trajectory tracking, optimizing bike-share distribution, and urban sensing.
The workshop participants discussed data access in
cities, data generating processes and inherent biases
in the collection protocols, challenges associated
with open city data, acquiring domain knowledge
and expertise, impacting and informing policy, and
finally, the need for a tighter integration of the AI
community and its subfields to make an impact on
urban science and city governance.
AI research has the opportunity to transform our
cities around the world for the better by helping
improve operations, services, security, citizen participation, and quality of life.
Theo Damoulas wrote this report and served as
chair of the event. The papers of the workshop were
published as AAAI Press Technical Report WS-15-04.
Artificial Intelligence for
Transportation: Advice, Interactivity,
and Actor Modeling
The transportation domain is increasingly taking up
AI techniques in the design of products and systems.
Cars nowadays implement machine-learning algo-
rithms. When searching for a route on a mobile app,
solutions are provided through AI algorithms. This
workshop has covered a diverse selection of topics,
both from a more theoretical and more practical
nature, showing once again that there are many
transportation problems where applying AI technol-
ogy is beneficial. Traffic signal control was featured
in two presentations, one of which was an invited
talk given by Scott Sanner, a principal researcher at
NICTA and the Australian National University.
The problem of fairly distributing the costs to a set
of clients is important, among other domains, in
vehicle-routing problems, where a fleet of vehicles
deliver goods to a set of clients. While Shapley cost
allocations are known to be optimal when a few relevant assumptions are imposed, the exact computation of such optimal allocations is a very challenging
computational problem. One presentation focused
on new hardness results, as well as a number of more
tractable, but not necessarily optimal techniques.
The increasing availability of car parking data
makes it possible to utilize AI algorithms for a more
effective use of car parking lots. More specific topics
covered in the workshop include pricing strategies
aimed at maximizing the occupancy of parking lots,
and predicting the occupancy of parking spots within a time window in the future.
In an electric vehicle, the amount of the battery
power available is a valuable resource. Battery power
is needed not only to power the engine, but also to
run auxiliary systems, such as the climate-control
system. Part of the work presented in the workshop
has focused on developing an adaptive-advise agent
that makes recommendations to the driver about the
settings of the climate-control system.
Navigating in a hostile environment brings up the
need to avoid, as much as possible, potential ambush
locations set by an adversary. This was the topic of
one presentation, which presented strategies for
planning routes optimized for ambush avoidance.
Aerial transportation was another well-represented topic. One presentation addressed the problem of
planning the trajectory of a helicopter or tilt rotor
craft while respecting a number of noise-related constraints. Another presentation focused on towing aircraft at an airport with self-driving vehicles.
Most work featured in the workshop was motivated by important real-world problems, evidence of
the important maturation of the community of
researchers working on the intersection of AI and
The last part of the workshop was an open discussion that revisited at a deeper level the topics presented and discussed earlier in the day. At the conclusion of the event, there was a consensus that
similar events should continue to be organized, possibly in colocation with major AI conferences.
Adi Botea and Sebastiaan Meijer wrote this report
and served as cochairs of the event. The papers presented in the workshop are available as AAAI Technical Report WS-15-05.
Beyond the Turing Test
The Turing test, now more than 60 years old, has
long served as a highly visible, public signpost for
research in artificial intelligence. But competitors
like Eugene Goostman and PARRY often seem like
exercises in evasion, rather than robust advances in