Page 24 - EE Times Europe Magazine - June 2025

24 EE|Times EUROPE

Pony AI Update: Robotaxis and Robotruck Services

Table 5: Pony AI Technology and Product Overview
Key Information Other Information
Integrated AI architecture based on LLM Virtual driving simulation: PonyWorld
E2E AV: From sensors to driving control Based on Nvidia Drive Orin system
AI-based AV tech
E2E simplifies AV system technologies Due to fewer connected AI models
Reinforced learning AI technology Key to improving virtual drivers
Software platform for driving AVs Proprietary software stack, algorithms
Controls all compute hardware and sensors Different robotaxi, robotruck versions
Virtual driver
Robotruck has higher-performance sensors Robotruck has longer-range sensor
Most important Pony AI product Used in all Pony AI AVs: OEMs and TNCs
PonyPilot Mobile app for Pony AI robotaxi users Reserve, pay, manage robotaxi rides
AV simulation system for virtual driver Validates virtual driver safety/reliability
PonyWorld Creates millions of driving testing scenarios Scenario testing to improve virtual driver
Improves virtual driver capability and safety More efficient than road testing
Seven LiDARs on top and sides of AVs Three long-range radars (4D)
Sensors
11 high-resolution cameras GNSS and IMU navigation/location sensorss
AV requests remote assistance RA operator monitors multiple AVs
Remote assistance (RA)
RA operator sends instructions to AV AV performs suggestion if safety is met
Three redundancy levels in the event Levels are normal, degraded safety,
Redundancy
of a system failure and minimal risk
(Source: Pony AI data; Compilation: Egil Juliussen, May 2025)

development and testing environment, significantly enhancing safety Sensor complement
metrics while reducing both testing time and associated costs. The Pony AI’s IPO document describes the sensors used in its sixth-
system answers various what-if questions and provides the ability to generation robotaxi. Figure 1, reproduced from that document, pro-
repeatedly test and make mistakes to improve the safety and efficiency vides good details on the sensors used in robotaxis.
of the AV system. The reinforcement learning technology is a key to The robotaxis deploy seven LiDARs on the top and sides of the
learning from AI-generated driving environments and decision-making current AV models to generate precise and real-time 3D images of the
experiences to refine AV performance. surrounding area, in a wide range of environments and under diverse

Figure 1: Pony AI robotaxi sensor system (Source: Pony AI IPO documentation)

JUNE 2025 | www.eetimes.eu