王亮 (Liang Wang)

Pos-doc Associate
Massachusetts Institute of Technology

CSAI Lab, Department of EECS
MIT, Cambridge MA 02139,
Office: 32-335
Phone: 617-324-8447
Email: wangliang@csail.mit.edu

Biography: Liang Wang was an undergraduate at Beijing Jiaotong University, Beijing. His Ph.D. was also from Beijing Jiaotong University. Now, he is a Pos-doc associate at MIT, supervised by Prof. Berthold K. P. Horn.

Curriculum Vitae: Curriculum Vitae

Divert & Alert:

In the United States, every year, several people whose job takes them to the sides of roads, are injured or killed by roadside collisions. Recently, this tragedy happened again. Massachusetts state trooper Thomas Clardy died when his cruiser was struck along the MA turnpike.
Poster of the Divert & Alert system - Machine vision is used to save the lives of roadside personnel
Real time headlamp detecting and tracking on the cell phone - video demo1 and video demo2
Generate and update a convex hull of given points - video demo
Detect the dangerous region automatically (i.e. update the convex hull according to the data)
See video demo1, video demo2 and video demo3
Finally, danger can be identified intelligently, and the warning signal will be sent promptly.
See video demo.
The whole system is runing on a cell phone in real time - video demo.
This work is summaried in our paper Machine Vision to Alert Roadside Personnel of Night Traffic Threats (Accepted by IEEE Trans. on Intelligent Transportation Systems).
See the data set and the experimental results (running on the cell phone).
The underlying idea for the "Divert and Alert" system is due to the late Prof. Seth Teller.

Fig. 1: In the United States, every year, several police officers are injured or killed by roadside collisions. Machine vision may save their lives.

Fig. 2： Generate and update the convex hull automatically.

Fig. 3: Using the convex-hull technique, the Normal Traffic Region (NTR) can be detected “intelligently”.

Bilateral Control:

Prof. Berthold K. P. Horn's pioneering work on bilateral control. The homepage of bilateral control (more simulations).
The complementary paper of bilateral control: Wave Equation of Suppressed Traffic Flow Instabilities (accepted to publish in IEEE Trans. on Intelligent Transportation Systems). In this paper, we prove the stability of bilateral control (comparing to car following model.).
In our new work on the stability analysis of bilateral control, We use a new approach - semi discrete analysis - to show the stability of bilateral control and instability of car-following model (under various boundary conditions), This work has been summaried in our new paper Eigenvalue and Eigenvector Analysis of Stability for a Line of Traffic (published in Applied Mathematics (STUD APPL MATH)). The MATLAB codes: code1 and code2
We build a simulation platform (see the Matlab code). Some interesting examples: example1, example2, example3 and example4. These simulations give some hint of collision free condition for bilateral control. The analysis is summarized in the paper COLLISION-FREE CONDITION ANALYSIS OF BILATERAL CONTROL MODEL (unsubmitted)
We also analyze the real implementation of bilateral control (red curves: analytical solution; black curves: numerical simulation). Then, we find stability conditions for various types of bilateral-control implementations. These results are summaried in the paper Stability Condition of Bilateral Control Strategies (unsubmitted)
We extend the simplest bilateral control (3-points) as a multi-node bilateral control model, which damped the low-frequency disterbation much faster. Thus, the system goes to the equilibrium state much more efficiently. These results are summaried in the paper Multi-node bilateral control model (submitted to IEEE Transactions on Automatic Control). Some simulation resultS (videos): simplest bilateral control (3-node), 15-node bilateral control with coefficients obtained by Taylor series approximation of -\omega^2 (BC-TS-7), 15-node bilateral control with coefficients obtained by least square approximation of -\omega^2 (BC-LSS-7), 15-node bilateral control with coefficients obtained by least square approximation of -|\omega| (BC-LSA-7), 15-node bilateral control with coefficients obtained by least square approximation of z(\omega) = min(-\omega^2,-|\omega|) (BC-LSZ-7). These are the simulations in the paper. The initial condition is: 40 seconds after car-following condtrol. See Fig. 6 and Fig. 7. The MATLAB code used for the simulation can be download here.

Fig. 4: Bilateral control eliminates traffic flow instability. See bilateral control.

Fig. 5: The simulation platform (See the Matlab code). When the black point drops below the blue dash, car colliion happens.

The simulation of multi-node bilateral condensed:

Fig. 6: Three node (1, -2, 1) bilateral control (First 40 seconds is controlled by car-following control) .

Fig. 7: BC-LSZ-7: 15-node bilateral control (First 40 seconds is controlled by car-following control).

Chain Stability of BCM:

For the bi-directional information flow system, e.g. bi-directional platooning or bilateral control, the stability property has been well studied.
However, at least to our knowledge, the string-stability analysis for the bi-directional information flow system is rarely studied.
Due to the specificity of the topological structure of the bi-directional system, the technic used for string-stability analysis of the one-directional information flow system can not be extend to analyze the string stability of the two directional information flow system directly.
Thus, new technic should be developed to deal with the ``string stability'' analysis for the bi-directional topological structure.
We prove that vehicles under bilateral control is (generalized) string stable, which is also called chain stable

Demo 1： In the beginning, all the vehicles are equally spaced by 25 meters, and move at the same speed 25 m/s. The simulation results, i.e. the trajectories of vehicles in space-time domain, is shows in a relative reference system, which is move at the constant speed 25 m/s. The initial space between the successive vehicle trajectories is 30 meters (i.e. initial vehicle space plus car length).

Demo 2: The black squares indicate the CFM vehicles, while the red squares indicate the BCM vehicles. The solid black square indicates the last car in the CFM-chain. The solid red square indicates the last car in the BCM-chain.

Shock Wave and Traffic flow:

Traffic flow - The nonlinear wave equation is derived from the conservation law. (Still working on this topic).
Shock waves - If shocks appear, what will we see? (Still working on this topic).

Traffic flow. The nonlinear wave equation is derived from the conservation law..

The traffic flow density is 1/(delta_x). This is the key to understand traffic problems.

Useful Math Tools:

Lax-Wendroff - the Finite Difference shceme with second order accuracy. How to extend the 3-points Lax-Wendroff scheme to N-points scheme? The key is trying to understand the finite difference scheme from the interpolation viewpoint. More information, better interpolation (e.g. Hermit interpolation).
Conservation Law and Shock waves - This is also known as the Initial value problem. One interesting book: Linear and Nonlinear Waves (by G.B. Whitham).
Another Way of teaching Calculus - one-line Determinism (By Prof. Qun Lin (林群)). How to teach junior high school students Calculus? Prof. Qun Lin will tell you the answer: Qun's big picture of calculus. 林群积分求高图 ——《人民日报》和《光明日报》（1997年）.

王 亮 (Liang Wang)

Pos-doc Associate Massachusetts Institute of Technology