Unsafe distance from the vehicle ahead is one of the leading causes of accidents in urban settings. Ask Ramesh, a trainer for commercial bus drivers.

Ramesh Ladukar works with a Nagpur City Bus operator that runs the Nagpur Aapli service. Ramesh trains his drivers in defensive driving practices. But he has found it difficult to enforce basic disciplines such as safe headway and lane discipline via only classroom sessions. Classroom learnings are easily forgotten, and drivers fall back to old behaviors. Training outcomes tend to be short-lived. A new approach to driver skilling is required.

But in the last three months, over 60% of his commercial drivers have been able to improve safe headway discipline. Ramesh is now able to track safety scores of his drivers. Not just that, he now knows which of his drivers are having trouble following basic driving disciplines and who needs personalized coaching.

What changed?

Trainers like Ramesh are key stakeholders in Project iRASTE, that has piloted Advanced Driver Assistance Systems (ADAS) in public transit vehicles. Ramesh and his team are pioneering the shift from classroom safety trainings to real-time driver safety scores and personalized coaching. With real-time in-cabin safety alerts, 60% drivers of Nagpur Aapli buses fitted with ADAS units have shown improvement in safe driving behaviours. This is a significant step towards governments commitment towards safer roads, while also initiating a major upgrade in driver skill trainings.

Classroom trainings to Continual Trainings: A new approach to driver skill trainings with ADAS

Traditionally, driver skill gap has been addressed via classroom trainings, test tracks or other spot checks in the field. Driver performance improves post the intervention but drops soon after. This is because such interventions are not undertaken regularly as it disrupts driver’s work schedule. Hence the only way to improve average driver performance remains hiring of experienced drivers. But for a fleet that sees high churn in its driver population, that is not an option. Its average driver performance never improves or worse, drops. And trainers like Ramesh don’t even realize the drop in performance till an unfortunate accident happens on road.

Advanced Driver Assistance Systems (ADAS) provide the opportunity to move away from Classroom training to Continual Training. With Continual training, instead of disrupting regular work schedule, coaching occurs on-the-move for both the trainer and his driver team. Drivers get real-time feedback with ADAS, instead of classroom instructions. In place of one-off assessments, trainers receive monthly safety scores for each driver providing regular reinforcement. Continual training thus makes coaching real-time, learner-centered and personalized, representing a major upgrade in driver skilling. This allows fleets to boost average driver performance, even while managing high churn in driver population.

Following are illustrations of Continual training on-the-move.

Drivers of ADAS-fitted buses receive real-time safety alerts. AI-powered Advanced Driver Assistance Systems (ADAS) continuously monitors the road ahead and warns the driver a few seconds before a potential collision.

Figure 1: ADAS safety alerts delivered to driver

ADAS alerts provide real-time warnings and quick feedback to drivers. Such real-time warnings improve driver reaction time by up to 2X. Thus, irrespective of driving experience of the driver, driver alertness to road events improves and likelihood of dangerous road incidents reduces.

Figure 2: Repeat warnings on unsafe distance ignored by the driver, resulting in a forward collision warning

Figure 3: Driver alerted on Pedestrian in Danger zone, resulting in a collision warning

Figure 4: Driver applying hard brake, alerted by a Forward collision warning

Trainers receive driver safety scores based on ADAS alerts. Every month, Ramesh uses driver scores to assess skill gap and create personalized coaching plans for drivers.

Figure 5: Illustration of driver performance on multiple ADAS alerts. In this instance, driver has shown notable improvement in scores.

What is the impact on fleet’s safety performance?

With one-off trainings, driver performance improves post the intervention but drops soon after. With Continual Training, the objective is to assess sustained improvement in safety performance of individual drivers. As previous approaches to trainings like classroom trainings suffer from improvements that are short-lived, sustained improvement is key when assessing the impact of Continual Training approach. Such sustained improvements can boost the average performance of a fleet’s driver pool significantly.

Study overview

As part of Project iRASTE, approximately one-third of Nagpur Aapli fleet was part of the pilot on Continual Training based on ADAS. When a driver is able to comprehend ADAS alerts and take action on the alerts, the driver is demonstrating improvement in defensive driving behaviors (such as safe headway distance, lane discipline, improved alertness to vulnerable road users, ie; pedestrians, 2-wheelers). Such improved behaviour on road can be measured by a reduction in ADAS alerts associated with that driver. Thus a reduction in ADAS alert count can reliably be considered as a leading indicator of drivers safety performance and is used to assess sustained improvements.

The initial pilot with 50 vehicles was conducted from Dec’21 – Mar’22. After learnings from this initial pilot and feedback from driver trainers, the broader deployment of ADAS units was undertaken in April-May, 2022.

  • 280 drivers were part of overall observation
  • 16 Lakh KMs of driving data, covering 150 NMC buses were analyzed (i.e., 1.6 million KMs)
  • Control period duration was 1 month
    • During Control period (CP), drivers do not receive any ADAS alerts in-cabin.
    • In this period, baseline performance of drivers is recorded
  • Observation period was designed to be 3 months. A longer observation period was chosen to analyze sustained performance
    • Observations were tabulated across 3 one-month periods – M1, M2, M3. During this period, drivers received in-cabin ADAS alerts
    • A large observation unit of 1 month was chosen to allow for sufficient randomness.
    • This minimizes the impact of a driver’s route assignment to driver scores

The results from this larger installation undertaken in April-May, 2022 are provided below.

Results

Figure 6: Summary of safety performance improvement of drivers across Control period (CP) & Observation periods (M1, M2, M3)

  • 60% drivers demonstrated sustained improvement in safe driving behavior at the end of observation periods
  • Collision risk reduced by ~40% for these improved drivers
  • Learning and adherence to ADAS alerts showed a drop in M2. Reinforcement trainings are key, esp in initial periods and helped to bring back improvement in M3.

ADAS Alert-wise observations

Figure 7: Summary of ADAS alert-wise performance of drivers. Forward Collision Warning (FCW) registered the largest decline

  • Forward Collision Warning (FCW) registered the largest decline (72% drivers improved), indicating a good decline in unsafe events on road
  • This was followed by Headway Monitoring and Warning (HMW) (60% drivers) and Lane Departure Warning (LDW) (63% drivers) – critical driving discipline for city drivers
  • Pedestrian Collision Warning (PCW) alerts are difficult to reduce in an urban setting. However, all operators have shown a positive trend in PCW alert reduction

Conclusion

Commercial vehicle trainers like Ramesh see high churn in driver population. With classroom trainings, fleets have sought to improve the number of good drivers and to identify skill gaps as soon as possible.  However, due to lack of regular reinforcement, drivers fall back to old behaviors. With previous approaches, average performance of a fleets driver pool thus stagnates or even drops.

As part of Project iRASTE, approximately one-third of Nagpur Aapli fleet was part of the pilot on Continual Training based on ADAS. 6 out of 10 drivers in this pilot have been able to demonstrate sustained improvement in safe driving behavior. Post such improvement, collision likelihood of ADAS-fitted buses is observed to be ~40% lower as compared to non-ADAS fitted buses.

Advanced Driver Assistance Systems (ADAS) provide the opportunity to move away from Classroom training to Continual Training. With Continual training based on ADAS, driver coaching gets real-time, learner-centric and personalized, representing a major upgrade in driver skilling. This allows fleets to boost average driver performance, even while managing high churn in driver population. Improvements in fundamental driving disciplines such as safe headway and lane discipline are note-worthy. Continual training based on ADAS provides a way to boost skill level of the entire fleet in a sustained manner.