Coordinated FMCW-OFDM ISAC Design

• Coordinated FMCW-OFDM is an integrated ISAC waveform that overlays FMCW radar chirps and OFDM data symbols over common time-frequency resources to support simultaneous sensing and communication.
• It leverages FMCW signals for precise target detection and channel estimation while allowing OFDM to carry payload data without dedicated pilot overhead.
• The design employs power-domain superposition and time-domain interference cancellation, yielding improved sensing accuracy and communication BER even under high Doppler.

Searching arXiv for the specific Co-FMCW-OFDM papers and closely related work to ground the article. arXiv_search(query="Coordinated FMCW OFDM integrated sensing communication", max_results=10) Coordinated FMCW-OFDM (Co-FMCW-OFDM) denotes an integrated sensing and communication (ISAC) transmission strategy in which a frequency modulated continuous-wave (FMCW) radar waveform and an orthogonal frequency division multiplexing (OFDM) communication waveform are deliberately coordinated and superimposed over the same time-frequency resources. In the formulations reported in the literature, the shared waveform is intended to support bistatic sensing and data transmission simultaneously, while reusing the same allocated bandwidth, RF front end, and antenna resources. A defining feature is that the FMCW component is not treated solely as a sensing signal: it is also exploited as a known reference for channel estimation and, in later formulations, for time-domain interference cancellation before OFDM demodulation (Sahin et al., 2020, Wang et al., 30 Sep 2025).

1. Conceptual basis and architectural scope

The central design principle of Co-FMCW-OFDM is non-orthogonal coexistence rather than orthogonal resource splitting. Instead of partitioning spectrum, time, or subcarriers between radar and communication, the waveform overlays FMCW and OFDM so that both functions reuse the same bandwidth. In the 2020 formulation, the system is described as a bistatic integrated sensing and communication scenario in which the transmitter and receiver are physically separated, and the receiver must both sense targets using radar-style measurements and decode OFDM communications (Sahin et al., 2020).

The later coordinated formulation makes the architectural commitment more explicit: the same RF front end, the same antenna(s), and the same spectrum are shared by both sensing and communication, and the FMCW and OFDM components are synchronized at symbol level (Wang et al., 30 Sep 2025). Within that architecture, the FMCW signal serves dual purposes. First, it is the sensing waveform for bistatic radar target detection. Second, it acts as a pilot-like known signal for channel estimation at the communication receiver.

This coordination distinguishes Co-FMCW-OFDM from several neighboring ISAC design classes cited in the literature. Some prior joint radar-communication systems use one waveform for both tasks, some allocate part of OFDM resources to sensing, some switch between radar and communication rather than supporting both simultaneously, and some analyze LFM radar and OFDM primarily as an interference coexistence problem rather than as a coordinated waveform design (Sahin et al., 2020). In contrast, Co-FMCW-OFDM treats coexistence as an intentional waveform-level coupling.

2. Waveform construction and frame organization

In the 2020 coexistence scheme, a single FMCW chirp sweeping linearly over bandwidth $\beta$ during duration $\tau$ is written as

$$s_{\text{chirp}}(t)=e^{j\pi\beta t^2/\tau}, \qquad 0\le t\le \tau.$$

The FMCW signal is a frame of $K$ chirps,

$$s_{\text{FMCW}}(t)=\sqrt{P_{\text{FMCW}}}\sum_{k=0}^{K-1} s_{\text{chirp}}(t-k\tau), \qquad 0\le t\le T,$$

while the OFDM waveform is formed from QAM symbols $\{d_n\}_{n=0}^{N-1}$ as

$$s_{\text{OFDM}}(t)=\sqrt{P_{\text{OFDM}}}\sum_{n=0}^{N-1} d_n e^{j2\pi n\Delta f t}, \qquad 0\le t\le T_s.$$

A cyclic prefix of duration $T_g$ is prepended to preserve orthogonality and convert linear convolution into circular convolution. After CP insertion, the full OFDM frame is denoted $\bar{s}_{\text{OFDM}}(t)$, with $T_{\text{OFDM}}=T_s+T_g$ (Sahin et al., 2020).

The transmitted baseband frame is then designed as

$\tau$0

Accordingly, the first chirp is radar-only, whereas the remaining interval carries the sum of FMCW and OFDM. The first interference-free chirp is used later for least-squares channel-coefficient estimation (Sahin et al., 2020).

The 2025 coordinated formulation preserves the superposition principle but changes the symbol organization. The OFDM communication waveform is written as

$\tau$1

with $\tau$2 and $\tau$3. The FMCW waveform is deliberately assigned an all-zero cyclic prefix, and its symbol duration is aligned with the OFDM symbol duration. After pulse shaping with $\tau$4, the transmitted signal is

$\tau$5

where $\tau$6 is the power allocated to OFDM and $\tau$7 is the power allocated to FMCW. This is a power-domain superposition, and the receiver observes both signals together rather than through separate radar and communication chains (Wang et al., 30 Sep 2025).

Taken together, these formulations show two concrete realizations of the same topic: an initial coexistence design based on a radar-only first chirp and a later symbol-synchronous design based on an all-zero FMCW CP.

3. Propagation model, bistatic operation, and the effective channel

The 2020 signal model adopts a bistatic, linear time-varying multipath channel with $\tau$8 targets or scatterers. The received passband signal is expressed as

$\tau$9

where $$s_{\text{chirp}}(t)=e^{j\pi\beta t^2/\tau}, \qquad 0\le t\le \tau.$$0 is the attenuation or reflection coefficient, $$s_{\text{chirp}}(t)=e^{j\pi\beta t^2/\tau}, \qquad 0\le t\le \tau.$$1 is the delay, $$s_{\text{chirp}}(t)=e^{j\pi\beta t^2/\tau}, \qquad 0\le t\le \tau.$$2 is the Doppler shift, $$s_{\text{chirp}}(t)=e^{j\pi\beta t^2/\tau}, \qquad 0\le t\le \tau.$$3 is the phase error, and $$s_{\text{chirp}}(t)=e^{j\pi\beta t^2/\tau}, \qquad 0\le t\le \tau.$$4 is AWGN. The large-scale path loss is given by

$$s_{\text{chirp}}(t)=e^{j\pi\beta t^2/\tau}, \qquad 0\le t\le \tau.$$5

with path-loss exponent $$s_{\text{chirp}}(t)=e^{j\pi\beta t^2/\tau}, \qquad 0\le t\le \tau.$$6, distance $$s_{\text{chirp}}(t)=e^{j\pi\beta t^2/\tau}, \qquad 0\le t\le \tau.$$7, and antenna gains $$s_{\text{chirp}}(t)=e^{j\pi\beta t^2/\tau}, \qquad 0\le t\le \tau.$$8 (Sahin et al., 2020).

After downconversion and sampling at $$s_{\text{chirp}}(t)=e^{j\pi\beta t^2/\tau}, \qquad 0\le t\le \tau.$$9, the discrete-time received signal becomes

$K$0

with

$K$1

This form makes the delay and Doppler dependence explicit, enabling subsequent FMCW-based estimation (Sahin et al., 2020).

The 2025 formulation introduces a more detailed received-signal model under arbitrary delays. Instead of constraining delays to integer multiples of the sampling period, it assumes

$K$2

With a raised-cosine pulse-shaping filter $K$3 of finite support $K$4, the sampled received signal becomes

$K$5

where

$K$6

This leads to an effective channel in which each physical target can expand into $K$7 effective paths, giving a total number of effective paths

$K$8

The corresponding effective-path parameters are

$K$9

with

$$s_{\text{FMCW}}(t)=\sqrt{P_{\text{FMCW}}}\sum_{k=0}^{K-1} s_{\text{chirp}}(t-k\tau), \qquad 0\le t\le T,$$0

This path-spreading model is central to the later coordinated formulation because it makes sensing, channel estimation, and interference cancellation substantially denser than in an integer-delay model (Wang et al., 30 Sep 2025).

4. Sensing mechanisms and radar-derived channel estimation

A defining characteristic of Co-FMCW-OFDM is that sensing outputs are reused for communication-side channel reconstruction. In the 2020 scheme, the receiver dechirps each chirp interval via

$$s_{\text{FMCW}}(t)=\sqrt{P_{\text{FMCW}}}\sum_{k=0}^{K-1} s_{\text{chirp}}(t-k\tau), \qquad 0\le t\le T,$$1

The dechirped data are arranged into a coherent processing interval matrix whose fast-time dimension corresponds to range bins and slow-time dimension corresponds to Doppler bins. A 2D FFT over this CPI matrix yields a range-Doppler map, and peaks identify $$s_{\text{FMCW}}(t)=\sqrt{P_{\text{FMCW}}}\sum_{k=0}^{K-1} s_{\text{chirp}}(t-k\tau), \qquad 0\le t\le T,$$2 for each target. Standard detection methods such as CFAR or STAP can be used to set detection thresholds, although thresholding is not the main focus (Sahin et al., 2020).

Once delay and Doppler are estimated, the radar-only first chirp is used to estimate the complex gains $$s_{\text{FMCW}}(t)=\sqrt{P_{\text{FMCW}}}\sum_{k=0}^{K-1} s_{\text{chirp}}(t-k\tau), \qquad 0\le t\le T,$$3 by least squares. Let

$$s_{\text{FMCW}}(t)=\sqrt{P_{\text{FMCW}}}\sum_{k=0}^{K-1} s_{\text{chirp}}(t-k\tau), \qquad 0\le t\le T,$$4

The coefficient vector is obtained as

$$s_{\text{FMCW}}(t)=\sqrt{P_{\text{FMCW}}}\sum_{k=0}^{K-1} s_{\text{chirp}}(t-k\tau), \qquad 0\le t\le T,$$5

with closed-form solution

$$s_{\text{FMCW}}(t)=\sqrt{P_{\text{FMCW}}}\sum_{k=0}^{K-1} s_{\text{chirp}}(t-k\tau), \qquad 0\le t\le T,$$6

Together with the estimated delays and Dopplers, this reconstructs the channel matrix $$s_{\text{FMCW}}(t)=\sqrt{P_{\text{FMCW}}}\sum_{k=0}^{K-1} s_{\text{chirp}}(t-k\tau), \qquad 0\le t\le T,$$7. Because the channel is estimated from the radar waveform itself, the system does not need dedicated OFDM pilots, and all OFDM subcarriers can carry data (Sahin et al., 2020).

The 2025 coordinated formulation develops two explicit low-complexity sensing algorithms. Fast cyclic correlation radar (FCCR) computes the cyclic correlation

$$s_{\text{FMCW}}(t)=\sqrt{P_{\text{FMCW}}}\sum_{k=0}^{K-1} s_{\text{chirp}}(t-k\tau), \qquad 0\le t\le T,$$8

followed by the slow-time FFT

$$s_{\text{FMCW}}(t)=\sqrt{P_{\text{FMCW}}}\sum_{k=0}^{K-1} s_{\text{chirp}}(t-k\tau), \qquad 0\le t\le T,$$9

so that $\{d_n\}_{n=0}^{N-1}$0 is the range-Doppler map. The method exploits the cyclic structure induced by the FMCW waveform and its all-zero CP, and its complexity is characterized by FFT-based operation counts:

$\{d_n\}_{n=0}^{N-1}$1

real multiplications and

$\{d_n\}_{n=0}^{N-1}$2

real additions (Wang et al., 30 Sep 2025).

Digital mixing and down-sampling (DMD) instead mimics conventional FMCW processing entirely in the digital domain. The receiver multiplies the received signal by the conjugate of the known FMCW waveform to form $\{d_n\}_{n=0}^{N-1}$3, low-pass filters the result to obtain $\{d_n\}_{n=0}^{N-1}$4, down-samples by factor $\{d_n\}_{n=0}^{N-1}$5, and then computes a column-wise FFT and row-wise IFFT:

$\{d_n\}_{n=0}^{N-1}$6

$\{d_n\}_{n=0}^{N-1}$7

The magnitude $\{d_n\}_{n=0}^{N-1}$8 yields the range-Doppler map. The desired term contains the beat frequency

$\{d_n\}_{n=0}^{N-1}$9

which carries delay information. The reported comparison is that FCCR is about $$s_{\text{OFDM}}(t)=\sqrt{P_{\text{OFDM}}}\sum_{n=0}^{N-1} d_n e^{j2\pi n\Delta f t}, \qquad 0\le t\le T_s.$$0 dB better than DMD in the simulated scenarios, whereas DMD is conceptually more hardware-friendly because it avoids analog matched filtering hardware and reduces sample rate after low-pass filtering (Wang et al., 30 Sep 2025).

5. Effective-channel reconstruction, SIC, and OFDM reception

In the later coordinated framework, sensing estimates are converted into an effective-channel estimate by successive interference cancellation (SIC). For the $$s_{\text{OFDM}}(t)=\sqrt{P_{\text{OFDM}}}\sum_{n=0}^{N-1} d_n e^{j2\pi n\Delta f t}, \qquad 0\le t\le T_s.$$1-th significant path, the reconstructed reference is

$$s_{\text{OFDM}}(t)=\sqrt{P_{\text{OFDM}}}\sum_{n=0}^{N-1} d_n e^{j2\pi n\Delta f t}, \qquad 0\le t\le T_s.$$2

The first-path coefficient estimate is

$$s_{\text{OFDM}}(t)=\sqrt{P_{\text{OFDM}}}\sum_{n=0}^{N-1} d_n e^{j2\pi n\Delta f t}, \qquad 0\le t\le T_s.$$3

and the reconstructed contribution is

$$s_{\text{OFDM}}(t)=\sqrt{P_{\text{OFDM}}}\sum_{n=0}^{N-1} d_n e^{j2\pi n\Delta f t}, \qquad 0\le t\le T_s.$$4

For the $$s_{\text{OFDM}}(t)=\sqrt{P_{\text{OFDM}}}\sum_{n=0}^{N-1} d_n e^{j2\pi n\Delta f t}, \qquad 0\le t\le T_s.$$5-th path, the recursion becomes

$$s_{\text{OFDM}}(t)=\sqrt{P_{\text{OFDM}}}\sum_{n=0}^{N-1} d_n e^{j2\pi n\Delta f t}, \qquad 0\le t\le T_s.$$6

with

$$s_{\text{OFDM}}(t)=\sqrt{P_{\text{OFDM}}}\sum_{n=0}^{N-1} d_n e^{j2\pi n\Delta f t}, \qquad 0\le t\le T_s.$$7

The reconstructed OFDM frequency response is then

$$s_{\text{OFDM}}(t)=\sqrt{P_{\text{OFDM}}}\sum_{n=0}^{N-1} d_n e^{j2\pi n\Delta f t}, \qquad 0\le t\le T_s.$$8

while the actual channel is

$$s_{\text{OFDM}}(t)=\sqrt{P_{\text{OFDM}}}\sum_{n=0}^{N-1} d_n e^{j2\pi n\Delta f t}, \qquad 0\le t\le T_s.$$9

The paper defines the channel NMSE as

$T_g$0

It also reports that the sensing-aided SIC channel estimator improves NMSE over a version without SIC (Wang et al., 30 Sep 2025).

The 2020 receiver uses the radar-derived channel estimate to cancel the FMCW component prior to communication demodulation:

$T_g$1

The OFDM receiver then applies CP removal and frequency-domain equalization. With CP insertion matrix $T_g$2 and CP removal matrix $T_g$3, the channel-frequency-response matrix is

$T_g$4

Its diagonal entries correspond to effective per-subcarrier gains, while off-diagonal terms arise from Doppler-induced intercarrier or inter-symbol effects. The OFDM symbols are recovered through a frequency-domain equalizer of the form

$T_g$5

where $T_g$6 (Sahin et al., 2020).

The 2025 formulation moves FMCW cancellation explicitly into the time domain before OFDM demodulation. The reconstructed FMCW component is

$T_g$7

and the interference-cancelled signal is

$T_g$8

Under perfect sensing and channel estimation, the FMCW terms cancel exactly, leaving only OFDM and noise. The paper emphasizes that this time-domain cancellation is more accurate than approaches based on averaged received signals or frequency-domain subtraction (Wang et al., 30 Sep 2025).

6. Performance, tradeoffs, and comparative position within ISAC

The 2020 paper reports that the proposed system achieves good sensing accuracy even if the signal-to-noise ratio is low, and that communication performance is only $T_g$9 dB less at the target bit-error rate of $\bar{s}_{\text{OFDM}}(t)$0 compared to the assumption of perfect channel state information without any pilot overhead over OFDM subcarriers. At $\bar{s}_{\text{OFDM}}(t)$1 dB, the range-Doppler map clearly identifies three targets. Radar and channel-estimation accuracy are summarized through the channel MSE

$\bar{s}_{\text{OFDM}}(t)$2

The communication evaluation assumes convolutional coding, interleaving to mitigate deep fading, and no pilot symbols on OFDM subcarriers (Sahin et al., 2020).

The 2025 study provides a more extensive set of comparative results. Its main simulation settings include $\bar{s}_{\text{OFDM}}(t)$3, $\bar{s}_{\text{OFDM}}(t)$4, $\bar{s}_{\text{OFDM}}(t)$5, $\bar{s}_{\text{OFDM}}(t)$6 kHz, bandwidth $\bar{s}_{\text{OFDM}}(t)$7 MHz, sampling rate $\bar{s}_{\text{OFDM}}(t)$8 MHz, carrier frequency $\bar{s}_{\text{OFDM}}(t)$9 GHz, QPSK, LDPC $T_{\text{OFDM}}=T_s+T_g$0 with code rate $T_{\text{OFDM}}=T_s+T_g$1, $T_{\text{OFDM}}=T_s+T_g$2, and $T_{\text{OFDM}}=T_s+T_g$3. Two targets are usually considered, with reflection powers $T_{\text{OFDM}}=T_s+T_g$4 dB and $T_{\text{OFDM}}=T_s+T_g$5 dB (Wang et al., 30 Sep 2025).

Within those simulations, FCCR consistently outperforms DMD by about $T_{\text{OFDM}}=T_s+T_g$6 dB in range and speed RMSE. At high SNR and low Doppler, range RMSE approaches about half of the theoretical range resolution

$T_{\text{OFDM}}=T_s+T_g$7

and the reported speed resolution is

$T_{\text{OFDM}}=T_s+T_g$8

Higher Doppler degrades range estimation for both algorithms. On the communication side, BER worsens as Doppler increases because OFDM orthogonality is increasingly disrupted, while interference cancellation helps substantially; “actual IC” is much better than “without IC,” although not as good as perfect cancellation because sensing and channel estimates are imperfect (Wang et al., 30 Sep 2025).

The power split $T_{\text{OFDM}}=T_s+T_g$9 operates as an explicit sensing-communication tradeoff parameter. Increasing $\tau$00 helps communication BER because more power goes to OFDM, but it hurts sensing and channel estimation because FMCW power is reduced. The paper defines the relevant power ratio as

$\tau$01

This identifies the power allocation between the FMCW and OFDM components as a tunable design knob rather than a fixed property of the waveform (Wang et al., 30 Sep 2025).

The coordinated scheme is also compared with conventional OFDM with embedded pilots and with a conventional OFDM-plus-FMCW approach. Against a 5G-like OFDM frame with DMRS pilots, Co-FMCW-OFDM yields significantly better range/Doppler sensing, better channel NMSE, and significantly improved BER, especially under high Doppler. The stated reasons are that conventional pilot structures are sparse and require interpolation in fast time-varying channels, while decision-feedback sensing degrades when decoded data are erroneous. Against the OFDM-plus-FMCW comparison, the coordinated formulation is reported to achieve lower uncoded BER even though the competing method assumes an ideal channel for interference cancellation. The stated reasons include the all-zero FMCW CP, the use of FMCW for both sensing and channel estimation, and time-domain cancellation before OFDM demodulation (Wang et al., 30 Sep 2025).

A common misconception is that superimposing FMCW and OFDM necessarily turns sensing into a communication burden or communication into a radar burden. The reported results do not support that simplification. In these formulations, the same FMCW component supplies range/Doppler information, enables channel estimation, and supports interference cancellation, while the OFDM component carries payload data without dedicated pilot tones in the 2020 scheme and with improved BER relative to pilot-based alternatives in the 2025 scheme. A plausible implication is that the technical difficulty is not the coexistence itself but the quality of sensing-informed channel reconstruction under Doppler, arbitrary delays, and imperfect cancellation.