From Feasibility to Desirability: Plan, Learn, Adapt (PLA) Framework for Personalized On-Device Itinerary Generation

2026-07-16

AI Summary

This work addresses the challenge of generating personalized travel itineraries that simultaneously satisfy combinatorial feasibility and subjective user preferences—a balance existing methods struggle to achieve, especially under mobile computational constraints. The authors propose Plan-Learn-Adapt (PLA), a three-stage framework that first constructs a diverse set of feasible candidate itineraries, then trains a lightweight reward model via pairwise comparisons to capture global preferences such as pacing and geographic coherence, and finally performs on-device local optimization while preserving feasibility. PLA uniquely decouples yet synergistically integrates feasibility enforcement and preference learning. Evaluated on 2,519 human assessments, it achieves a 67.8% win rate (an 11.2-percentage-point improvement over the best baseline) with 100% feasibility. The reward model demonstrates 67.6% cross-city generalization accuracy, and real-world deployment increases itinerary completion rates by 91%, with an average on-device latency of only 109.9 milliseconds.

Abstract

Generating personalized trip itineraries is a complex planning task and involves a tension between hard combinatorial feasibility and soft latent desirability. Classical optimization enforces constraints but fails to capture subjective traveler preferences. While learning-based approaches model preferences, they cannot guarantee feasibility. Mobile deployment imposes additional resource constraints on both. To address this, we propose Plan, Learn, Adapt (PLA), a three-stage framework for personalized on-device itinerary generation. The Plan stage builds a heterogeneous ensemble of lightweight planners that produces structurally diverse feasible candidates. From pairwise itinerary comparisons, Learn fits a compact Bradley-Terry reward model that captures emergent schedule properties such as pacing, geographic coherence, and day balance, which per-POI signals miss. Finally, Adapt applies feasibility-preserving local refinement within a device-aware compute budget; every intermediate state is feasible by construction. On 2,519 pairwise human comparisons across more than 100 U.S. cities, the reward-guided ensemble achieves a 67.8% win rate, 11.2 percentage points above the best single planner, with 100% feasibility. Three frontier LLMs, GPT-5, Claude Opus 4.5, and Gemini 3 Pro, achieve 0% feasibility under the same constraints. The reward model generalizes across held-out cities, with a 67.6% mean leave-one-city-out accuracy. In production deployment within FlyEnJoy, PLA increased itinerary completion rates by 91%, with 109.9 ms average on-device latency.

Problem

• personalized itinerary generation
• feasibility
• desirability
• on-device planning
• traveler preferences

Innovation

• on-device itinerary generation
• feasibility-preserving optimization
• Bradley-Terry reward model
• personalized trip planning
• lightweight ensemble planning