Conditional Handovers (CHO) in 5G Networks

Updated 2 January 2026

Conditional Handovers (CHOs) are a mobility mechanism that separates handover preparation and execution to ensure timely and reliable network transitions.
CHO significantly reduces interruptions, mobility failures, and signaling overhead through methods like FCHO and adaptive offsets.
CHO integrates learning-enabled control and dynamic resource optimization, offering robust performance in both terrestrial and non-terrestrial 5G deployments.

Conditional Handovers (CHOs) constitute a pivotal 3GPP mobility innovation for 5G and 5G-Advanced networks, enabling robust handover execution under rapid channel dynamics and diverse deployment regimes. Unlike legacy deterministic handover schemes, CHO explicitly decouples the handover process into a resource-intensive preparation phase and a deferred, measurement-conditioned execution phase, allowing user equipment (UE) to locally trigger execution when predefined criteria are met. This paradigm enhances reliability, delivers substantial reductions in mobility failures and outages, and, through recent enhancements, supports adaptable architectures and learning-based control in both terrestrial and vertical (e.g., NTN) scenarios.

1. Technical Principles and Protocol Flow

The CHO mechanism, standardized in 3GPP Release 16, divides handover into two separate phases:

Preparation Phase: The UE monitors Layer-3 filtered RSRP (Reference Signal Received Power) from serving ( $c_0$ ) and candidate ( $c'$ ) cells, applying a preparation offset $O_{\mathrm{prep}}$ over a timer $T_{\mathrm{prep}}$ . If

$P_{c'}^{\mathrm{L3}}(m) > P_{c_0}^{\mathrm{L3}}(m) + O_{\mathrm{prep}}, \quad\text{for } m \in [m_0 - T_{\mathrm{prep}}, m_0],$

is satisfied, the UE transmits an RRC measurement report. The serving gNB responds by provisioning up to eight candidate target cells (multi-candidate CHO) and exchanges Handover Request/Ack messages to reserve resources at the targets.

Execution Phase: The UE continually checks for the execution condition, usually with a lower offset $O_{\mathrm{exec}}$ over $T_{\mathrm{exec}}$ :

$P_{c'}^{\mathrm{L3}}(m) > P_{c_0}^{\mathrm{L3}}(m) + O_{\mathrm{exec}}, \quad\text{for } m \in [m_1 - T_{\mathrm{exec}}, m_1].$

Upon satisfaction, the UE executes the stored handover command locally, with no further network signaling. Execution can involve Contention-Free or Contention-Based Random Access (CFRA/CBRA), depending on preallocated resources (Iqbal et al., 2024, Stanczak et al., 2022).

The preparation and execution offsets ( $O_{\mathrm{prep}}, O_{\mathrm{exec}}$ ) and their respective timers critically determine the trade-off between fast response to channel fluctuations and robustness against premature handovers or outages (Deb et al., 2024).

The release and replace mechanisms allow the network to withdraw preparations for a candidate that becomes obsolete or swap in a stronger new candidate if the UE’s prepared list is at capacity (Iqbal et al., 2022).

State-of-the-art extensions, such as Fast Conditional Handover (FCHO), allow UEs to retain and reuse earlier preparations after a successful handover, autonomously executing subsequent handovers without requiring re-preparation, which further reduces signaling overhead (Iqbal et al., 2022, Iqbal et al., 2022).

2. Performance Analysis and Key Metrics

Performance of CHO is rigorously quantified in terms of:

Handover Interruption Time ( $T_\mathrm{HO}$ ): In standard CHO, interruption is dominated by PRACH-related steps and fine timing acquisition, typically around 54 ms; early timing advance acquisition (RACH-less CHO) reduces this to around 30 ms, yielding a 43.2% reduction in interruption time and an 18.7% reduction in outage (Iqbal et al., 2024).
Mobility Failure Probability ( $P_{\mathrm{fail}}$ ): Defined as the ratio of failed handovers (HOF) and radio link failures (RLF) to total attempts. For example, FCHO achieves reduction of mobility failures by $10.5\%$ at $60\,\mathrm{km/h}$ and $19.3\%$ at $120\,\mathrm{km/h}$ , even under worst-case hand blockage (Iqbal et al., 2022). Analytical Markov models explicitly relate $P_{\mathrm{fail}}$ to $O_{\mathrm{prep}}, O_{\mathrm{exec}}, T_{\mathrm{prep}}, T_{\mathrm{exec}}$ , channel fading (Rayleigh, Rician), and UE speed (Deb et al., 2024).
Packet Loss and Handover Latency: Mean handover latency $L_{\mathrm{HO}}$ and associated packet loss $P_{\mathrm{loss}}$ are given in closed form as:

$L_{\mathrm{HO}} = T_{\mathrm{sample}}\,\left(\frac{n}{p_{\mathrm{prep}}} + \frac{m}{p_{\mathrm{exec}}}\right), \qquad P_{\mathrm{loss}} = \lambda L_{\mathrm{HO}},$

where $p_{\mathrm{prep}}, p_{\mathrm{exec}}$ are probabilistic success rates, $n,m$ are the numbers of samples for preparation and execution windows, respectively, and $\lambda$ is the downlink packet arrival rate (Deb et al., 2024).

Signaling Overhead: Standard CHO incurs high signaling load through numerous preparation, release, and replacement messages. FCHO reduces CHO signaling by approximately 27–36% (Iqbal et al., 2022, Iqbal et al., 2022).

These quantitative metrics are validated through both system-level simulation (e.g., 420 UEs in 7-site wrap-around networks) and analytic Markov models.

3. Random Access, Beamforming, and Resource Management

CHO operations in FR2 and mmWave beamformed networks involve complex interactions with the RACH and Random Access Channel (RACH) resource management:

Contention-Free Random Access (CFRA): CHO prefers CFRA for handover execution due to its reduced collision probability and deterministic delay. However, because of the potentially long interval between preparation and execution, originally prepared CFRA beams may become faded. To address this, recent proposals introduce beam-specific measurement reporting and CFRA updating, where UEs signal if a non-CFRA beam outperforms the prepared beam, prompting an updated resource allocation (Stanczak et al., 2023). This process leads to a reduction in contention-based random access by up to 13% and a handover-delay decrease of up to 10 ms, at the cost of modest signaling overhead.
Learning-Augmented RACH Assignment: Supervised decision-tree learning (BELL) is applied to classify the root cause of HO failures (coverage hole, wrong beam, early execution, wrong cell), and adjusts beam-specific preparation offsets accordingly, which can yield up to 60% reduction in handover failures associated with beam errors (Karabulut et al., 2019, Iqbal et al., 2023).
Resource-Efficient RACH (RE-RACH): Forcing CFRA on any prepared beam—regardless of access threshold SNR—maximizes success probability and minimizes RACH-induced interruption time and signaling (Karabulut et al., 2019).

4. Adaptive Control, Fast Conditional Handover, and Resource Optimization

Fast Conditional Handover (FCHO): FCHO is proposed for further reducing signaling and latency by allowing the UE to keep the configurations of previously prepared target cells after a handover, enabling extremely rapid subsequent handover execution. This technique reduces median resource reservation time slightly (from 0.47 s to 0.40 s under optimized block/offset schemes) and signaling overhead by up to 35%, with negligible loss to outage or mobility robustness (Iqbal et al., 2022, Iqbal et al., 2022).
Resource Reservation Optimization: Techniques including “block listing” (excluding highly improbable handover target cells), reduction of preparation offsets for weak candidates, and combination approaches minimize resource holding time and signaling while maintaining or slightly improving mobility performance (Iqbal et al., 2022).
Parameter Selection: The choice of $O_{\mathrm{prep}}, O_{\mathrm{exec}}, T_{\mathrm{prep}}, T_{\mathrm{exec}}$ must be made conditional on the fading environment and UE velocity. Larger offsets and shorter timers are recommended under high Doppler/Rayleigh conditions to avoid oscillations and ensure robustness, with typical values: $O_{\rm{prep}}=5$ –$10$ dB, $O_{\rm{exec}}=1$ –$3$ dB, $T_{\rm{prep}}=80$ –$100$ ms, $T_{\rm{exec}}=60$ –$80$ ms (Deb et al., 2024).

5. Algorithmic and Meta-Learning Frameworks for CHO

Meta-Learning-Based Orchestration: With increased architectural complexity, frameworks such as CONTRA and CHOMET leverage meta-learning on near-real-time RIC (Radio Intelligent Controller) platforms to jointly optimize traditional handovers and conditional handovers (Kalntis et al., 10 Jul 2025, Kalntis et al., 26 Dec 2025). These methods employ multi-expert online learning, optimize dynamic regret versus oracle (full-future) comparator sequences, and adaptively select HO type per-UE, balancing throughput, switching cost, and signaling.
- CONTRA and CHOMET show at least 180% utility gain (dynamic regret reduction) and up to 89.5% lower regret versus 3GPP CHO baselines in volatile conditions (Kalntis et al., 10 Jul 2025, Kalntis et al., 26 Dec 2025).
- The action set comprises the set of prepared cells per UE per slot, with per-slot utility functions encoding rate, signaling/resource use, and switching costs.
- Real-world and simulation studies confirm throughput improvements and lower cumulative handover cost in dense, highly variable environments.

6. Advanced Use Cases, Recovery, and Future Directions

Non-Terrestrial Networks (NTN), Integrated Access Backhaul (IAB), NR-U: CHO protocols are extended to support timing-based and location-based execution conditions (e.g., during satellite handover or when parent migration triggers in IAB). Recovery procedures, where the UE exploits previously stored CHO configurations to accelerate reestablishment after handover failure, achieve recovery rates above 80% under suitable parameters (Stanczak et al., 2022).
Analytical and System Guidelines:
- For cell-edge or severe fading: prefer higher offsets, shorter timers.
- For mild fading or low velocity: longer timers, smaller offsets suffice.
- SON (self-organizing network) loops are recommended for continuous adaptation of mobility parameters to match real-time network and mobility statistics (Deb et al., 2024).
Research and Standardization Directions:
- Multi-slice, multi-RAT control, federated meta-learning.
- Integration of beam/trajectory predictions and dual-connectivity.
- Enhancements for resource pooling, rapid CFRA handovers, and further reduction in interruption times toward the sub-millisecond regime.

7. Quantitative Comparison and Configuration Guidelines

Scheme	Mobility Failure Rate Reduction	Signaling Overhead Reduction	Outage Reduction	Reference
FCHO vs. CHO (120 km/h)	19.3%	27%	18.7%	(Iqbal et al., 2022, Iqbal et al., 2024)
RACH-less CHO	–	+27% (radio), +20% (Xn)	43.2% (int. time), 18.7%	(Iqbal et al., 2024)
CHOMET/CONTRA vs. 3GPP	–	~25%	≥180% utility gain	(Kalntis et al., 10 Jul 2025, Kalntis et al., 26 Dec 2025)

Recommended network configurations for robust CHO depend on velocity, fading, and deployment architecture. Preparing 2–4 candidates with appropriately chosen $O_{\mathrm{prep}}, O_{\mathrm{exec}}$ and dynamic meta-learning-based orchestration achieves optimal mobility robustness, signaling cost, and user-rate performance across diverse 5G and beyond deployment scenarios.