Driver Guidance System (DGS) for Taxi Drivers

DGS enables taxi drivers to be more productive by digesting real-time demands and competitions.

Background

Uber and Uber-like services have been rocking the taxi industry globally for the past few years. Traditional taxi companies have been slow in responding to these challenges. The development of the DGS technology (Jha et al., 2018; Jha et al., 2018) aims to help taxi drivers to compete against ride-hailing technologies by providing guidance to an area with predicted demand (i.e., actionable driving decision) that would help drivers shorten their vacant cruising time before finding the next passenger.

The key enabling technology behind the development of the DGS is the stream data processing, demand prediction (Cheng & Rathnayaka, 2025; Cheng & Rathnayaka, 2023), and personalized decision support system (Varakantham et al., 2012; Ahmed et al., 2012).

Figure 1: The Architecture of DGS.

The DGS App

The technology is delivered via a smartphone App. The DGS App is designed to be hands-free; as drivers drive around, the App will automatically provide the best recommendations for individual drivers depending on their respective locations.

Figure 2: The DGS App.

Simulated Performance Evaluation

The expected performance of DGS App is evaluated a massive scale agent-based simulation (Cheng & Nguyen, 2011). As shown in Figure 3 below, the personalized guidance ensures that DGS stays competitive even when the adoption ratio increases.

Figure 3: DGS performance (average trips) in response to adoption ratio.

Field Trial Results

The DGS is field tested in Singapore from 2017 to 2018. In total we have recruited 500 taxi drivers to try DGS. On average, following DGS guidance results in 34% reduction in vacant roaming time (Cheng et al., 2018).

The reduction in the vacant roaming time is particularly evident when the demand levels are low (e.g., from 11pm to 5am in Figure 4). We also track the performance of DGS across Singapore, and the improvements are universal (Figure 5).

Figure 4: The performance (average vacant roaming time) of DGS vs. non-DGS trips at different hours of the day. The bars at the bottom correspond to the demand level. The value of guidance is particularly evident when the demands are low.

Figure 5: The performance (average vacant roaming time) of DGS vs. non-DGS trips at different regions across Singapore.

Finally, we also breakdown the performance of DGS by demand channels such as “street hail” or “booking”. We can see that while the booking channel indeed helps drivers in reducing their vacant roaming time (from ~13min to 9min), having personalized guidance still helps in further bringing down the vacant roaming time by ~24%.

Furthermore, after accounting for the “response time” (i.e., the time between driver’s commitment to the job and the passenger pick-up), we can see that the street-hail channel with DGS can be even more efficient than the booking channel (35% vacant vs. 43% vacant).

Figure 6: The performance (average vacant roaming time) of DGS vs. non-DGS trips at different regions across Singapore.

The DGS has also been commercially tested in Tokyo, Japan in April 2020. With dedicated taxi drivers hired to follow DGS exclusively, we observe that DGS-guided drivers experienced 12% less vacant time than non-DGS drivers.

References

2025

Patent

Method and System for Taxi Demand Prediction Using a Neural Network Model

Shih-Fen Cheng, and Prabod Rathnayaka

Singapore Patent 10202103115Q, Mar 2025

PDF

2023

IEEE BigData-23
M^2-CNN: A macro-micro model for taxi demand prediction

Shih-Fen Cheng, and Prabod Rathnayaka

In 2023 IEEE International Conference on Big Data, Dec 2023

Abs Bib PDF

In this paper, we introduce a macro-micro model for predicting taxi demands. Our model is a composite deep learning model that integrates multiple views. Our network design specifically incorporates the spatial and temporal dependency of taxi or ride-hailing demand, unlike previous papers that also utilize deep learning models. In addition, we propose a hybrid of Long Short-Term Memory Networks and Temporal Convolutional Networks that incorporates real world time series with long sequences. Finally, we introduce a microscopic component that attempts to extract insights revealed by roaming vacant taxis. In our study, we demonstrate that our approach is competitive against a large array of approaches from the literature on the basis of detailed moving logs of more than 20,000 taxis and 12 million trips per month over a three-month period. Our analysis of the effectiveness of individual components reveals that microscopic information is essential for generating high-quality predictions.
@inproceedings{cheng_bigdata_2023, title = {M$^{2}$-CNN: A macro-micro model for taxi demand prediction}, booktitle = {2023 IEEE International Conference on Big Data}, address = {Sorrento, Italy}, author = {Cheng, Shih-Fen and Rathnayaka, Prabod}, year = {2023}, month = dec, }

2018

IAAI-18
Upping the game of taxi driving in the age of Uber

Shashi Shekhar Jha, Shih-Fen Cheng, Meghna Lowalekar, and 6 more authors

In Thirtieth Annual Conference on Innovative Applications of Artificial Intelligence, Feb 2018

Abs Bib PDF

In most cities, taxis play an important role in providing point-to-point transportation service. If the taxi service is reliable, responsive, and cost-effective, past studies show that taxi-like services can be a viable choice in replacing a significant amount of private cars. However, making taxi services efficient is extremely challenging, mainly due to the fact that taxi drivers are self-interested and they operate with only local information. Although past research has demonstrated how recommendation systems could potentially help taxi drivers in improving their performance, most of these efforts are not feasible in practice. This is mostly due to the lack of both the comprehensive data coverage and an efficient recommendation engine that can scale to tens of thousands of drivers. In this paper, we propose a comprehensive working platform called the Driver Guidance System (DGS). With real-time citywide taxi data provided by our collaborator in Singapore, we demonstrate how we can combine real-time data analytics and large-scale optimization to create a guidance system that can potentially benefit tens of thousands of taxi drivers. Via a realistic agent-based simulation, we demonstrate that drivers following DGS can significantly improve their performance over ordinary drivers, regardless of the adoption ratios. We have concluded our system designing and building and have recently entered the field trial phase.
@inproceedings{jha_iaai_2018, title = {Upping the game of taxi driving in the age of Uber}, booktitle = {Thirtieth Annual Conference on Innovative Applications of Artificial Intelligence}, address = {New Orleans, Louisiana, USA}, author = {Jha, Shashi Shekhar and Cheng, Shih-Fen and Lowalekar, Meghna and Wong, Nicholas and Rajendram, Rishikeshan and Khiem, Tran Trong and Varakantham, Pradeep and Nghia, Truong Trong and Rahman, Firmansyah}, year = {2018}, month = feb, }
AAMAS-18
A driver guidance system for taxis in Singapore

Shashi Shekhar Jha, Shih-Fen Cheng, Rishikeshan Rajendram, and 5 more authors

In Seventeenth International Conference on Autonomous Agents and Multiagent Systems, May 2018

Best Application Demo Award Abs Bib PDF

Awarded Best Application Demo Award

Traditional taxi fleet operators world-over have been facing intense competitions from various ride-hailing services such as Uber and Grab.Based on our studies on the taxi industry in Singapore, we see that the emergence of Uber and Grab in the ride-hailing market has greatly impacted the taxi industry: the average daily taxi ridership for the past two years has been falling continuously, by close to 20% in total. In this work, we discuss how efficient real-time data analytics and large-scale multiagent optimization technology could help taxi drivers compete against more technologically advanced service platforms. Our system has been in field trial with close to 400 drivers, and our initial results show that by following our recommendations, drivers on average save 21.5% on roaming time.
@inproceedings{jha_aamas_2018, author = {Jha, Shashi Shekhar and Cheng, Shih-Fen and Rajendram, Rishikeshan and Wong, Nicholas and Rahman, Firmansyah Bin Abd and Trong, Nghia Truong and Lowalekar, Meghna and and Pradeep Varakantham}, title = {A driver guidance system for taxis in Singapore}, booktitle = {Seventeenth International Conference on Autonomous Agents and Multiagent Systems}, address = {Stockholm, Sweden}, year = {2018}, month = may, }
AAMAS-18
Taxis strike back: A field trial of the driver guidance system

Shih-Fen Cheng, Shashi Shekhar Jha, and Rishikeshan Rajendram

In Seventeenth International Conference on Autonomous Agents and Multiagent Systems, May 2018

Abs Bib PDF

Traditional taxi fleet operators world-over have been facing intense competitions from various ride-hailing services such as Uber and Grab (specific to the Southeast Asia region). Based on our studies on the taxi industry in Singapore, we see that the emergence of Uber and Grab in the ride-hailing market has greatly impacted the taxi industry: the average daily taxi ridership for the past two years has been falling continuously, by close to 20% in total. In this work, we discuss how efficient real-time data analytics and large-scale multi-agent optimization technology could potentially help taxi drivers compete against more technologically advanced service platforms.
@inproceedings{cheng_aamas_2018, title = {Taxis strike back: A field trial of the driver guidance system}, booktitle = {Seventeenth International Conference on Autonomous Agents and Multiagent Systems}, address = {Stockholm, Sweden}, author = {Cheng, Shih-Fen and Jha, Shashi Shekhar and Rajendram, Rishikeshan}, year = {2018}, month = may, }

2012

AAAI-12
Decision support for agent populations in uncertain and congested environments

Pradeep Varakantham, Shih-Fen Cheng, Geoff Gordon, and 1 more author

In Twenty-Sixth AAAI Conference on Artificial Intelligence, Feb 2012

Abs Bib PDF

This research is motivated by large scale problems in urban transportation and labor mobility where there is congestion for resources and uncertainty in movement. In such domains, even though the individual agents do not have an identity of their own and do not explicitly interact with other agents, they effect other agents. While there has been much research in handling such implicit effects, it has primarily assumed deterministic movements of agents. We address the issue of decision support for individual agents that are identical and have involuntary movements in dynamic environments. For instance, in a taxi fleet serving a city, when a taxi is hired by a customer, its movements are uncontrolled and depend on (a) the customers requirement; and (b) the location of other taxis in the fleet. Towards addressing decision support in such problems, we make two key contributions: (a) A framework to represent the decision problem for selfish individuals in a dynamic population, where there is transitional uncertainty (involuntary movements); and (b) Two techniques (Fictitious Play for Symmetric Agent Populations, FP-SAP and Softmax based Flow Update, SMFU) that converge to equilibrium solutions.We show that our techniques (apart from providing equilibrium strategies) outperform "driver" strategies with respect to overall availability of taxis and the revenue obtained by the taxi drivers. We demonstrate this on a real world data set with 8,000 taxis and 83 zones (representing the entire area of Singapore).
@inproceedings{varakantham_aaai_2012, author = {Varakantham, Pradeep and Cheng, Shih-Fen and Gordon, Geoff and Ahmed, Asrar}, title = {Decision support for agent populations in uncertain and congested environments}, booktitle = {Twenty-Sixth AAAI Conference on Artificial Intelligence}, address = {Toronto, Canada}, year = {2012}, month = feb, }
UAI-12
Uncertain congestion games with assorted human agent populations

Asrar Ahmed, Pradeep Varakantham, and Shih-Fen Cheng

In Twenty-Eighth Conference on Uncertainty in Artificial Intelligence, Aug 2012

Abs Bib PDF

Congestion games model a wide variety of real-world resource congestion problems, such as selfish network routing, traffic route guidance in congested areas, taxi fleet optimization and crowd movement in busy areas. However, existing research in congestion games assumes: (a) deterministic movement of agents between resources; and (b) perfect rationality (i.e. maximizing their own expected value) of all agents. Such assumptions are not reasonable in dynamic domains where decision support has to be provided to humans. For instance, in optimizing the performance of a taxi fleet serving a city, movement of taxis can be involuntary or nondeterministic (decided by the specific customer who hires the taxi) and more importantly, taxi drivers may not follow advice provided by the decision support system (due to bounded rationality of humans). To that end, we contribute: (a) a general framework for representing congestion games under uncertainty for populations with assorted notions of rationality. (b) a scalable approach for solving the decision problem for perfectly rational agents which are in the mix with boundedly rational agents; and (c) a detailed evaluation on a synthetic and real world data set to illustrate the usefulness of our new approach with respect to key social welfare metrics in the context of an assorted human-agent population. An interesting result from our experiments on a real-world taxi fleet optimization problem is that it is better (in terms of revenue and operational efficiency) for taxi drivers to follow perfectly rational strategies irrespective of the percentage of drivers not following the advice.
@inproceedings{ahmed_uai_2012, author = {Ahmed, Asrar and Varakantham, Pradeep and Cheng, Shih-Fen}, title = {Uncertain congestion games with assorted human agent populations}, booktitle = {Twenty-Eighth Conference on Uncertainty in Artificial Intelligence}, address = {California, USA}, year = {2012}, month = aug, }

2011

IAT-11
TaxiSim: A multiagent simulation platform for evaluating taxi fleet operations

Shih-Fen Cheng, and Thi Duong Nguyen

In 2011 IEEE/WIC/ACM International Conference on Intelligent Agent Technology, Aug 2011

Abs Bib PDF

Taxi service is an important mode of public transportation in most metropolitan areas since it provides door-to-door convenience in the public domain. Unfortunately, despite all the convenience taxis bring, taxi ﬂeets are also extremely inefficient to the point that over 50% of its operation time could be spent in idling state. Improving taxi ﬂeet operation is an extremely challenging problem, not just because of its scale, but also due to fact that taxi drivers are self-interested agents that cannot be controlled centrally. To facilitate the study of such complex and decentralized system, we propose to construct a multiagent simulation platform that would allow researchers to investigate interactions among taxis and to evaluate the impact of implementing certain management policies. The major contribution of our work is the incorporation of our analysis on the real-world driver’s behaviors. Despite the fact that taxi drivers are selﬁsh and unpredictable, by analyzing a huge GPS dataset collected from a major taxi ﬂeet operator, we are able to clearly demonstrate that driver’s movements are closely related to the relative attractiveness of neighboring regions. By applying this insight, we are able to design a background agent movement strategy that generates aggregate performance patterns that are very similar to the real-world ones. Finally, we demonstrate the value of such system with a real-world case study.
@inproceedings{cheng_iat_2011, author = {Cheng, Shih-Fen and Nguyen, Thi Duong}, title = {TaxiSim: A multiagent simulation platform for evaluating taxi fleet operations}, booktitle = {2011 IEEE/WIC/ACM International Conference on Intelligent Agent Technology}, address = {Lyon, France}, year = {2011}, month = aug, }