Road traffic congestion is inevitable when road infrastructure fails to keep pace with
increasing traffic demand, especially on urban roads. Adaptive traffic signals possess
better potential than conventional signals to alleviate congestion and ensure uninterrupted traffic flow. Compared to countries with homogeneous and lane-disciplined
traffic, adaptive signals on Indian roads have had limited success, mainly due to traffic’s
heterogeneous and lane-less nature (termed as ‘mixed traffic’ in this research). There
is also a scarcity of associated literature in this field. Among the few reported studies,
those based on model-based control theory are limited, where mathematical models
represent the physics of the system, resulting in less effective capturing of real-time
traffic characteristics. Based on this motivation, this thesis attempts to solve the research
problem of developing appropriate model-based adaptive traffic signal control schemes
to reduce congestion within an over-saturated network in mixed traffic conditions
with traffic density, a good indicator of congestion, to characterize the traffic under
consideration.The study concentrates on the research challenges of developing suitable mathematicalmodels and estimating and controlling density under mixed traffic conditions.

In the first stage of the research, two macroscopic non-continuum models represented
in state-space form were developed based on the vehicle conservation principle. In the
first model, the width of the road stretch under study was divided into multiple parallel
strips to incorporate lane indiscipline, and a composition-based weighted vehicle length
was used to incorporate heterogeneity. The second model was developed by considering
the entire road width as a single unit to address lane indiscipline, and area occupancy,
which considers the varying vehicle dimension of the traffic, was used to incorporate
heterogeneity. Two dynamic model-based density estimation schemes based on the
strip-based approach and area occupancy-based approach using the Kalman filtering
technique were developed and were corroborated using simulated data. Performance
was tested for different traffic scenarios such as congestion and non-recurrent traffic
incidents. To improve the estimation accuracy in scenarios involving transitions in
traffic conditions, two adaptive estimators were developed. In the next stage, an adaptive
model-based control scheme to maintain optimal density in a link under over-saturated
conditions was developed using the area occupancy model mentioned above. Since the
basic model-based state feedback control algorithm was inadequate for a traffic network
with multiple intersections and constraints, a model predictive control (MPC) scheme
was designed. The control scheme integrated with the estimator was implemented in
a VISSIM-MATLAB simulation environment. Performance evaluation showed a mean
absolute percentage error (MAPE) of less than 4 % in tracking the desired traffic density,
indicating very good performance of the developed MPC scheme.