TrafficPal

Safer, more efficient traffic signals.

  • Find the best timing plans and road changes with TrafficPal AI
  • Test them in a digital twin of your road network

Retiming today

  • Calculated, never tested. No one runs the plan under realistic traffic.
  • Simulation is out of budget. A calibrated model costs more than the project can carry.
  • Fine-tuned on live traffic. Adjusted in the field, on your hours.
  • Bigger changes are guesswork. Phasing, detectors or lanes are judged by experience.
  • Safety is checked by hand. Conflicts show up only on the street.
  • Hard to show the client. A timing sheet doesn’t show what changes, or why.

Retiming with TrafficPal AI

TrafficPal AI on West 26th Street, Erie. An example study, in simulation.

How it works

  1. Upload your data

    Nothing to model yourself

    • Turn counts
    • Controller configurations
    • Intersection geometry

  2. We build your digital twin

    Built by us in a few days

    • Your roads, and every detector and signal controller as configured in the field
    • Traffic from your data, calibrated against your real counts
    • In the browser, nothing to install

  3. Put TrafficPal AI to work on it

    Walk away. Come back to tested plans.

    • State your goal and constraints; it proposes changes and tests each one in simulation against what runs today
    • Come back to tested plans, every change listed for you to inspect, adjust or undo
    • The data, a report and a page to share with stakeholders, yours to keep

What TrafficPal AI does for your team

TrafficPal AI reads your network, proposes changes and tests each one in simulation against what runs today. The digital twin is yours without TrafficPal AI too: test any plan of your own the same way.

A retiming study, step by step

Here is what each step of a study asks of you, with the tools most firms use today and with TrafficPal.

Step Today’s tools signal-timing and simulation software TrafficPal
Building the model
Weak. Weeks of specialist time. Lanes, volumes and timings entered by hand, intersection by intersection; a simulation also needs its controllers coded and calibrated. All on a desktop install.
Strong. Delivered ready, in your browser. Upload your data and get access in days, not weeks. Nothing to install; open it from anywhere.
Finding a plan
Weak. Optimized by formula. Plans are searched with delay equations and one fixed objective; a simulation, if there is one, only checks the plan they pick.
Strong. An AI engineer searches with you. You set the goal and the limits; it proposes plans and refines them from each result. Step in at any time: the final choice is yours.
Testing and comparing
Weak. Rarely simulated. A simulation is usually out of budget; when there is one, each plan is run on its own and their report tables compared afterwards.
Strong. Every plan simulated, all on the same plots. Against the same calibrated traffic as today’s plan: travel times, queues, conflicts and the signal performance measures you already use, side by side.
Lanes, phasing, detectors
Partial. Possible, by hand. Each new lane or phasing is coded by hand, then estimated by formula or simulated again.
Strong. Changed like a timing plan. You or the AI engineer change them and test them against your goal.
Safety
Weak. Not built in. At best, vehicle paths are exported to separate tools.
Strong. In every run. Pedestrian and vehicle conflicts, and a check against the MUTCD and the agency’s rules.
For the client
Partial. Static reports. Printed analysis tables; shared online dashboards, where offered, are an add-on.
Strong. A results web page, not just a report. Send the client a page with the comparisons, their plots and clips of the simulation; a printable report too.

From public documentation for the most widely used signal-timing and traffic-simulation software, and FHWA’s Traffic Analysis Toolbox, October 2026.

Everything else it does

  • Thinks with you Ask it why while it works, and it answers with the evidence. Step in, or stop it, at any time.
  • Follows your rules Every plan is checked against the MUTCD and your state's and your client agency's rules.
  • Never locked out What it makes is an ordinary plan: every change listed, editable by hand, or undone.
  • Your call When it is a judgment call, it lays out the trade-off and asks. The last word is yours.
Trial B vs today’s plan, in simulation Standards check passed
Travel time
9:40 to 8:30
Stops per vehicle
2.4 to 1.9
Pedestrian conflicts
7 to 7
Vehicle conflicts
23 to 19
Portrait of Ozan Tonguz

Ozan Tonguz

Founder and CEO, VTL, a Carnegie Mellon spin-off Professor of Electrical and Computer Engineering, Carnegie Mellon University

Ozan’s research spans vehicular and wireless networks and artificial intelligence; before Carnegie Mellon, he did research at Bell Communications Research (Bellcore). Today he leads a team bringing research at the edge of applied AI into the daily work of traffic engineers.

  • 300+ IEEE papers
  • 21 patents
  • Work covered in 100+ countries

LinkedIn Carnegie Mellon profile

Your expertise, multiplied.

Priced per study. Early access is open now.

Early access

We’re taking on a few first firms. We build the twin of a corridor you choose and deliver it at an early-access price, and you tell us what works and what doesn’t. Places are limited.

Talk to our engineers

Ask about the product, early access or your next study. We reply by email.

We use your details only to reply to you. Privacy notice